defeating-control-flow-flattening
Defeats control-flow-flattening obfuscation by identifying the dispatcher/state- variable structure and reconstructing the original control flow so the logic becomes readable. Activates for requests t
SKILL.mdskills/defeating-control-flow-flattening/SKILL.md
GitHubDefeating Control-Flow Flattening
When to Use
- A function is flattened (a dispatcher loop switching on a state variable, with original blocks as switch cases) and you need to recover the original control flow.
- You are reversing OLLVM-style or similar flattening.
Do not use this on non-flattened code — confirm the dispatcher/state-variable pattern first. This skill analyzes disassembly/CFG data statically and executes nothing.
Prerequisites
- A disassembled function's basic-block/edge data (e.g., exported from Ghidra/IDA/Binary Ninja as JSON).
Workflow
Step 1: Detect the flattening structure
python scripts/analyst.py detect cfg.json
Identifies a dispatcher block (high in-degree, switch/compare on a state variable) and the relay/ case blocks that all branch back to it.
Step 2: Recover state transitions
Track the constant assigned to the state variable in each case to determine the successor block, rebuilding the true edges.
Step 3: Reconstruct the original CFG
Emit the deflattened successor mapping (case → next case) for re-annotation in the disassembler.
Step 4: Verify
Confirm the recovered flow has a single entry, sane successors, and no orphan blocks.
Validation
- The dispatcher is identified by in-degree and state-variable comparison.
- Each case's next-state constant maps to a real successor block.
- The reconstructed CFG covers all original (non-relay) blocks.
Pitfalls
- State variable computed (not a constant) — requires emulation/symbolic execution.
- Multiple dispatchers or nested flattening.
- Opaque predicates and bogus blocks inflating the case set.
References
- See
references/api-reference.mdfor the deflattener. - OLLVM and ATT&CK T1027 references (linked in frontmatter).