binary-lifting

name	binary-lifting
description	Expertise in binary lifting techniques - converting machine code to LLVM IR for analysis, decompilation, and recompilation. Use this skill when working on reverse engineering, binary analysis, deobfuscation, or converting binaries to higher-level representations.

Binary Lifting Skill

This skill covers techniques and tools for lifting binary executables to LLVM IR, enabling advanced analysis, transformation, and recompilation of existing binaries.

Core Concepts

What is Binary Lifting?

Binary lifting is the process of translating low-level machine code (x86, ARM, etc.) into a higher-level intermediate representation (LLVM IR), enabling:

• Static and dynamic analysis
• Deobfuscation and vulnerability research
• Code recompilation and optimization
• Cross-architecture translation

Lifting Pipeline

Binary → Disassembly → IR Generation → Optimization → Analysis/Recompilation

Major Lifting Frameworks

Production-Grade Tools

• RetDec (Avast): Full decompiler with C output, multi-architecture support
• McSema (Trail of Bits): x86/x64 to LLVM IR, function recovery
• revng: Based on QEMU, supports multiple architectures
• reopt (Galois): Focus on correctness and formal methods

Research/Specialized Tools

• Rellume: Fast x86-64 to LLVM lifting for JIT scenarios
• fcd: Pattern-based decompiler with optimization passes
• bin2llvm: QEMU-based binary to LLVM translator
• llvm-mctoll: Microsoft's machine code to LLVM lifter

Language-Specific Lifters

• llvm2c/IR->C: Convert LLVM IR back to C code
• llvm2cranelift: LLVM IR to Cranelift IR
• Leaven: LLVM IR to Go language
• masxinlingvonta: JVM bytecode to LLVM IR

Implementation Techniques

Instruction Semantics Translation

// Example: Translating x86 ADD to LLVM IR
Value* translateADD(IRBuilder<> &builder, Value* op1, Value* op2) {
    Value* result = builder.CreateAdd(op1, op2, "add_result");
    
    // Update flags (CF, OF, SF, ZF, etc.)
    updateCarryFlag(builder, op1, op2, result);
    updateOverflowFlag(builder, op1, op2, result);
    updateSignFlag(builder, result);
    updateZeroFlag(builder, result);
    
    return result;
}

Control Flow Recovery

1. Linear Sweep: Simple but misses code with embedded data
2. Recursive Descent: Follow control flow, better coverage
3. Speculative Disassembly: Handle indirect jumps/calls
4. Machine Learning: Use ML to identify function boundaries

Handling Indirect Control Flow

• Value Set Analysis (VSA)
• Symbolic execution for jump target resolution
• Type recovery for virtual table reconstruction

Triton Integration

Triton symbolic execution engine can be used with lifting:

from triton import TritonContext, ARCH, Instruction

ctx = TritonContext(ARCH.X86_64)

# Symbolically execute and extract AST
inst = Instruction(b"\x48\x01\xd8")  # add rax, rbx
ctx.processing(inst)

# Convert Triton AST to LLVM IR
ast = ctx.getRegisterAst(ctx.registers.rax)
llvm_ir = triton_ast_to_llvm(ast)

Deobfuscation via Lifting

Approach

1. Lift obfuscated binary to LLVM IR
2. Apply optimization passes to simplify
3. Use custom passes for specific obfuscation patterns
4. Re-emit cleaned code

Useful Optimization Passes

• Dead Store Elimination (DSE)
• Global Value Numbering (GVN)
• Constant Propagation
• Instruction Combining
• Loop Simplification

VMP/VM Handler Recovery

• Identify dispatcher patterns
• Extract VM bytecode semantics
• Convert handlers to native IR
• Example: TicklingVMProtect for VMProtect analysis

Best Practices

1. Architecture Support: Handle endianness, calling conventions, ABI differences
2. Memory Modeling: Accurate memory layout for global/stack variables
3. External Dependencies: Handle library calls and system calls
4. Validation: Compare execution traces of original vs lifted code
5. Incremental Lifting: Support partial program analysis

Dynamic Binary Lifting

Runtime Translation

• Instrew: Fast instrumentation through LLVM
• QBDI: QuarkslaB Dynamic Binary Instrumentation
• binopt: Runtime optimization of binary code

JIT Recompilation

Lift frequently executed code paths for runtime optimization:

• Profile-guided lifting
• Hot path detection
• Speculative optimization

Resources

For a complete list of lifting tools and research papers, refer to the LIFT section in the main README.md.

Getting Detailed Information

When you need detailed and up-to-date resource links, tool lists, or project references, fetch the latest data from:

https://raw.githubusercontent.com/gmh5225/awesome-llvm-security/refs/heads/main/README.md

This README contains comprehensive curated lists of:

• Binary lifting frameworks and tools (LIFT section)
• Related research papers and documentation
• Implementation examples and tutorials