Abstract

Code obfuscation has become a vital tool to protect, for example, intellectual property against competitors. In general, it attempts to impede program understanding by making the to-be-protected program more complex. As a consequence, a human analyst reasoning about the obfuscated code has to overcome this barrier by transforming it into a representation that is easier to understand.

In this training, we get to know state-of-the-art code obfuscation techniques and look at how these complicate reverse engineering. Afterwards, we gradually familiarize ourselves with different deobfuscation techniques and use them to break obfuscation schemes in hands-on sessions. Thereby, participants will deepen their knowledge of program analysis and learn when and how (not) to use different techniques.

First, we have a look at important code obfuscation techniques and discuss how to attack them. Afterwards, we analyze a virtual machine-based (VM-based) obfuscation scheme, learn about VM hardening techniques and how to tackle them.

In the second part, we cover SMT-based program analysis. In detail, students learn how to solve program analysis problems with SMT solvers, how to prove characteristics of code, how to deobfuscate mixed Boolean-Arithmetic and how to break weak cryptography.

Before we use symbolic execution to automate large parts of code deobfuscation, we first introduce intermediate languages and compiler optimizations to simplify industrial-grade obfuscation schemes. Following, we use symbolic execution to automate SMT-based program analysis and break opaque predicates. Finally, we learn how to write disassemblers for virtualization-based obfuscators and how to reconstruct the original code.

The last part covers program synthesis, an approach to simplify code based on its semantic behavior. After collecting input-output pairs from binary code, we not only learn how to simplify large expression trees, but also how we can verify the correctness of simplifications. Then, we use program synthesis to deobfuscate mixed Boolean-Arithmetic and learn the semantics of VM instruction handlers.

Teaching

Note that the training focuses on hands-on sessions. While some lecture parts provide an understanding of when to use which method, various hands-on sessions teach how to use them to build custom purpose tools for one-off problems. The trainer actively supports the students to successfully solve the given tasks. After a task is completed, we discuss different solutions in class. Furthermore, students receive detailed reference solutions that can be used during and after the course.

While the hands-on sessions use x86 assembly, all tools and techniques can also be applied to other architectures such as MIPS, PPC or ARM.

Class Outline

The training orientates at the following outline:

• Introduction to Code (De)obfuscation

  • Motivation
  • Application scenarios
  • Program analysis techniques

• Code Obfuscation Techniques

  • Opaque predicates
  • Control-flow flattening
  • Mixed Boolean-Arithmetic
  • Virtual machines
  • Virtual machine hardening

• Code Deobfuscation Techniques

  • Compiler optimizations
  • Reconstructing control flow
  • SMT-based program analysis
  • Taint analysis
  • Symbolic execution
  • Program synthesis

• Compiler Optimizations

  • Dead code elimination
  • Constant propagation/folding
  • Static single assignment (SSA)
  • Optimizing obfuscated code

• SMT-based Program Analysis

  • SAT and SMT solvers
  • Encoding programs analysis problems for SMT solvers
  • Proving semantic equivalence
  • Proving properties of a piece of code
  • Solving complex program constraints
  • Deobfuscating mixed Boolean-Arithmetic
  • Breaking weak cryptography

• Symbolic Execution

  • Intermediate languages for reverse engineering
  • Symbolic and semantic simplification of obfuscated code
  • Automation in reverse engineering
  • Identifying virtual machine components
  • Interaction with SMT solvers
  • Breaking opaque predicates
  • Writing disassemblers for virtualization-based obfuscators

• Program Synthesis

  • Concept of program synthesis
  • Learning code semantics based on its input/output behavior
  • Obtaining input/output pairs from code
  • Methods to simplify large expression trees
  • Proving the correctness of simplifications
  • Deobfuscating mixed Boolean-Arithmetic
  • Learning semantics of VM instruction handlers

Requirements and Recommendations

Prerequisites

The participants should have basic reverse engineering skills. Furthermore, they should be familiar with x86 assembly and Python.

Software Requirements

Students should have access to a computer with 4 GB RAM (minimum) and at least 20 GB disk space. Furthermore, they should install a disassembler of their choice (e.g., IDA or Ghidra) as well as virtualization software such as Virtual Box or VMware. Students will be provided with a Linux VM containing all necessary tools and setups.