ecc-audit-engine

Reproducible secp256k1 correctness, differential-testing, failure-minimization, and dynamic-trace research engine.

• BigUint reference mathematics for secp256k1 laboratory oracles
• Deterministic structured corpus generation
• Differential testing against local C ABI targets
• Failure minimization and reproducer replay
• Valgrind Lackey dynamic instruction / memory-address tracing
• Synthetic calibration fixtures (branch, table-address, control)
• Bounded public-API baseline against pinned upstream libsecp256k1

• Not a wallet, exchange, or production crypto library
• Not a private-key cracker, “Bitcoin breaker”, or exploit engine
• Not a universal proof of constant-time behavior
• Not authorization to test third-party systems

• Synthetic private keys only
• Locally compiled binaries and authorized laboratory targets only
• No live wallets, exchanges, nodes, browser extensions, hardware wallets, or production services

No vulnerability in upstream libsecp256k1 has been confirmed by the published experiments.

Synthetic corrupted and leaky targets are calibration fixtures used to validate the engine.

A negative bounded trace result is not a universal proof of constant-time behavior.

git clone https://github.com/Yudis-bit/ecc-audit-engine.git
cd ecc-audit-engine
./scripts/bootstrap.sh
./scripts/verify.sh

verify.sh checks prerequisites, builds, lints, tests, runs differential and minimizer gates, synthetic dynamic-trace calibration (when Valgrind is available), pinned libsecp baseline, schema validation, and writes a compact report under reports/readiness-run/.

cargo fmt --all --check
cargo clippy --workspace --all-targets --all-features -- -D warnings
cargo test --workspace --all-features --locked -- --nocapture
./scripts/build_targets.sh
./scripts/run_differential.sh
./scripts/run_minimizer_replay.sh

export PATH="$PWD/third_party/valgrind/bin:$PATH"   # if using project-local Valgrind
./scripts/run_trace_calibration.sh
# full campaigns: scripts/run_dynamic_trace.sh

Manual CI workflow: .github/workflows/extended-trace.yml

See docs/dependencies.md.

Pinned upstream commit (historical v0.1.0 / reproducibility):
11dad6d06c0ea8fd6d9d423d32bddd18b70b8b53 (bitcoin-core/secp256k1)

See targets-src/SECP256K1_PIN.txt and docs/upstream-research.md.

• reports/published/ — compact public summaries
• reports/readiness-run/ — verification outputs (raw/generated gitignored)
• reports/dynamic-trace/ — backend metadata and calibration summaries
• Schemas: schemas/*-v1.schema.json

1. Implement the ABI in crates/target-api/include/ecc_target.h
2. Add sources under harnesses/
3. Extend scripts/build_targets.sh
4. Add corpus cases / experiments as needed
5. Run ./scripts/verify.sh

cargo run -p cli -- differential \
  --target targets/corrupted-target.so \
  --corpus fixtures/corpus-v1.json \
  --case <case_id> \
  --minimize \
  --output reports/latest

For dynamic-trace campaigns, see experiments/dynamic-trace/ and scripts/run_trace_calibration.sh.

See SECURITY.md and docs/threat-model.md. Do not publish exploitable detail for unmodified upstream without coordinated disclosure.

• Architecture
• Threat model
• Experiment methodology
• Report schema
• Reproducibility
• Upstream research
• Limitations
• Dependencies
• Release process

MIT — see LICENSE.