RE-BREAKER

Offense-research toolkit for defeating anti-reverse-engineering protection on binaries the analyst has the legal right to analyze.

RE-BREAKER builds on RE-AI v2.9.2 (the analyzer + fingerprinter + IL lifter) and adds the explicit bypass primitives that RE-AI's charter forbids. If you need to lift encrypted method bodies from a binary you have the right to RE, this is the tool. If you're trying to pirate software, go elsewhere — this isn't for you, and the legal exposure is on you.

Charter

This is a defensive + offensive research toolkit. The intended users are:

  • Reverse-engineering consultants (with a signed engagement letter for the target binary)
  • Malware analysts (with authorization to analyze the malware sample)
  • Blue-team / purple-team staff (analyzing detection rules + tooling)
  • Security researchers (publishing under responsible-disclosure)
  • Academic RE labs (peer-reviewed research)

The out-of-scope uses (which the LICENSE-OFFENSIVE.md clauses explicitly forbid):

  • Piracy of commercial software
  • Bypassing DRM to access content without paying
  • Unauthorized access to systems the analyst doesn't own
  • Any use that violates the laws of the operator's jurisdiction

If you accept that the re-dump CLI's --license-acknowledge flag requires, then you're operating within the charter. If you don't, the CLI refuses to run.

Relationship to the other repos

Repo Charter License What it does
RE-AI Identify, lift, fingerprint MIT The static analysis foundation. 31 MCP servers + 29 skills. Vendor-neutral in shipped content.
RE-Library Public docs on RE techniques MIT 12-category Astro site + PyPI MCP server. DRM-system names only in drm/.
RE-UNLEASHED Cite-only vendor attribution MIT Per-publisher / per-engine / per-protection doc tree. Names vendors but vendoring-forbids source.
RE-BREAKER (this) Offense research: bypass AGPL-3.0 + offensive-research-use clause The new bypass toolset. The tiered injection tool. The combined defender+offender technique catalog.

RE-BREAKER is self-contained as of v0.4.0: it no longer requires RE-AI to be checked out as a sibling. The RE-AI code RE-BREAKER depends on (PE parsing, anti-analysis correlation, Speakeasy emulation, rizin, YARA, capa, PDB) is vendored under vendored/re-ai/. See docs/ARCHITECTURE.md for the full layout.

What this v0.4.0 ships

  • 15 MCP servers in servers/ (was 12 in v0.3.0; +3 new in v0.4.0):
    • The 12 from v0.3.0: re-triage, re-il2cpp-triage, re-catalog-match, re-anti-debug-patch, re-patch-apply, re-runtime-dump, re-encrypted-vm-bypass, re-vm-decrypt, re-frida-runtime, re-c-injection-build, re-anti-vm-spoof, re-vendor-anti-tamper
    • NEW re-frida-wine-runtime — In-process frida-gadget injection (the only known-working Frida path on this host). 6 tools + 6 per-Pattern hook templates.
    • NEW re-injection-runtime — C-injection runtime (no Frida). 4 tools + 4 per-hook C source specs.
    • NEW re-winedbg — Wine + winedbg + gdb + GEF wrapper. 30 tools (core 10 implemented; GEF + convenience are stubs).
  • 19 skills in skills/ (per-Pattern + per-vendor bypass playbooks; unchanged in v0.4.0)
  • 8 vendored RE-AI servers under vendored/re-ai/servers/ (library code; not exposed as MCP):
    • re-lief, re-patch, re-anti-analysis, re-speakeasy, re-rizin, re-yara, re-capa, re-pdb
  • 7 vendored honest-read triage JSONs at Vendored RE-AI catalog data and triage outputs.
  • The combined defender+offender technique catalog (data/catalog.json, 59 entries — v0.4.0 NEW: entitlement family)
  • The YARA export of the catalog (data/yara/techniques.yar)
  • The 7 attack-pattern playbooks (docs/PLAYBOOKS/)
  • The C/C++ in-process DLL/SO injector (inject/src/)
  • The shared src/re_breaker/triage.py helper (replaces 7 near-identical _load_triage() functions)
  • The LICENSE-OFFENSIVE.md clause document
  • The new docs/ARCHITECTURE.md and docs/WINE.md

Critical bug fixes in v0.4.0:

  • A1: re-patch-apply no longer silently returns sites_patched: 0
  • A2: re-c-injection-build now builds both .so and .dll (the .dll build previously failed on POSIX mkdir(path, 0755))
  • B1: re-catalog-match no longer returns 0 matches for IL2CPP targets (flattens nested {launcher_*, GameAssembly_dll}.{RDTSC, ...} shape)
  • B2: Plan-only servers (re-anti-debug-patch, re-vm-decrypt, re-anti-vm-spoof, re-runtime-dump) no longer fail with "no triage.json found"

Quickstart

git clone https://github.com/Heretek-RE/RE-BREAKER.git
cd RE-BREAKER
python -m venv .venv-re-breaker
source .venv-re-breaker/bin/activate
pip install -e .

# Read the offensive-research-use clause (required before first run)
cat LICENSE-OFFENSIVE.md

# Run the catalog matcher on a target
re-catalog-match --target=/path/to/foo.exe --intent=both

# Run the tiered injection CLI
re-dump --target=/path/to/foo.exe --mode=emulator --output=/tmp/dump/ --license-acknowledge

License

  • AGPL-3.0-or-later for the code (see LICENSE).
  • Offensive-research-use clause as a separate LICENSE-OFFENSIVE.md that the CLI cats on first run and requires explicit acknowledgement.
  • Per the cite-only contract (inherited from RE-UNLEASHED), no third-party source is vendored, even when the upstream license permits it.

v0.4.1+

Bypass toolset hardening and runtime execution fixes. See CHANGELOG.md or the git log for details.

Roadmap

  • Per-Pattern bypass orchestrator
  • Per-vendor bypass skills reaching production coverage
  • Additional runtime execution paths (Speakeasy emulator, Frida attach, C-injection)