Obfuscation Passes

For Developers

The items in this section describe the different obfuscation passes available in O-MVLL. The documentation of the passes provides information about when and how to use the obfuscation as well as its limitations.

There is also an overview of the pass synthesized by the following items:

Protection
  • Static: the pass aims at protecting against static code analysis (decompilation, disassemblers, …)
  • Dynamic: the pass aims at protecting against dynamic code analysis (hooking, instrumentation, debugging, …)
Resilience

This property tries to give a level of strength against a fully automated attack. If such an attack exists, it is mentioned in the References section of the pass.

Obfuscation Overhead

The kind of overhead introduced in the program when using the given obfuscation pass. The values can be:

  • Code size: the assembly code is larger.
  • Data size: the raw data of the binary is larger.
  • Memory size: the size of the binary in memory is larger 1.
  • Execution: the execution can be slowed down.
Public Attacks

Whether there exist public attacks that could be applied to the current obfuscation, it is important to remember that, even though an attack may be put in place, it does not necessarily imply that the attack will scale and can be blindly applied to the protected binary.

For Reverse Engineers

First and foremost, none of these obfuscation passes is unbreakable, nor are they meant to be. That said, we warmly welcome new attacks on these obfuscation techniques that could result in a complete scalable deobfuscation.

We are aware that simple mixed boolean-arithmetic expressions (MBA) may be simplified with open-source tools like mrphrazer/msynth. Thus, given the Arithmetic Obfuscation pass applied to several real functions, could we manage to identify and create a pipeline to simplify the expressions?

If there is a new way to address this problem, we will love to reference this work in the associated documentation. If there also is an idea to circumvent your attack, feel free to open a PR or an issue that describes this idea.


  1. Uninitialized static values do not take space in the binary (.bss) compared to initialized values (.data). That’s why this item differs from Data size↩︎