Introduction

In this blog post, I will show what was my approach to statically extract indicators from these cobaltstrike_shellcode.exe executables.

As seen below, there are a lot of these executables in VirusTotal.

Reversing engineering the executable

The behavior of these files is very simple to understand as they just decrypt a large blob of data (which is the cobalt strike beacon) and transfer execution to it.

These executables start with the popular jmp/call/pop technique to start executing the decryption stub.

The call instruction seen at the bottom of the previous image will place the address of the next instruction on the stack, which in this case will be a pointer to data that is crucial to the decryption routine.

As seen below, the first DWORD is the initial XOR key used in the decryption routine, the second DWORD is a value that when it is XOR’d with the first DWORD will give out the size of the encrypted data and the third DWORD is here the encrypted data starts.

The decryption routine will basically loop through the encrypted data and XOR every 4 bytes with the key.

The first 4 encrypted bytes are decrypted using the hardcoded XOR key while the next will be decrypted with the previous 4 decrypted bytes. Hope that makes sense :)

To test the decryption routine I made a quick IDA script.

So now at this point, I only needed to automate the extraction of the encrypted beacon payload and try to extract the config from it.

But how could I achieve this statically? Well, the solution that I found to be more reliable was to develop a Yara signature that was good enough to catch the main shellcode body in a data set of more or less 100 samples and automate the extraction of the embedded beacon.

Below it’s possible to see my Yara rule matching the whole data set.

Now it was just a matter of using it in a python script to get the offset to the shellcode, extract the data required to emulate the decryption routine, and dump the decrypted cobalt strike beacon.

Now let’s try to parse the beacon config with the Sentinel One extractor.

As seen, it worked!

Note: Sometimes the beacons may have the “MZ” magic bytes stripped but that will be ok as long as the Yara rule keeps matching on the shellcode.

Now with a working script to decrypt the beacon and a working beacon config extractor, I can combine them to extract indicators automatically.

That’s it. The full script can be found in this gist.

Cheers!