Only the latest 3.x patch receives security fixes. Upgrade with:

pip install --upgrade sc-neurocore

If you discover a security vulnerability in SC-NeuroCore, please report it responsibly:

1. GitHub Security Advisories (preferred): Report a vulnerability
2. Email: protoscience@anulum.li — Subject: [SECURITY] SC-NeuroCore — <brief description>
3. Do not open a public GitHub issue for security vulnerabilities.

We will acknowledge receipt within 48 hours and aim to provide a fix within 7 days for critical issues.

SC-NeuroCore is a neuromorphic computing simulation library. It does not handle user authentication, financial data, or network services in its default configuration. Security concerns are primarily:

• Malicious input files (JSON configs, NumPy .npz, ONNX models)
• Unsafe deserialization (pickle, NumPy load)
• Numerical overflow / denial of service via pathological inputs
• Native code memory safety (Rust engine via PyO3)
• HDL generation safety (SystemVerilog emitter input validation)
• Supply chain integrity (dependency audit)

The v3 Rust engine uses safe Rust by default. All unsafe blocks are limited to SIMD intrinsics with documented invariants. PyO3 boundary types enforce correct Python ↔ Rust conversions.

• Pickle allowlist restricts deserialization to known-safe classes
• Path traversal prevention on file I/O operations
• Array shape/dtype validation on all public API entry points

• Rust: cargo audit in CI
• Python: Minimal dependencies (numpy, scipy). No pickle.load of untrusted data in any production module.
• SBOM/release artefacts: run the offline audit locally with python tools/supply_chain_audit.py. Use --strict before releases to fail on SBOM metadata drift as well as hard errors such as un-hashed release requirements.

• No fuzzing harness yet. Property-based testing via proptest (Rust) and hypothesis (Python) covers many input paths, but dedicated fuzzing has not been set up.
• No third-party security audit. The codebase has not been reviewed by an external security firm.
• No CVE history. No vulnerabilities have been reported to date.

Contributions to improve security coverage (fuzzing harnesses, static analysis integration, audit reports) are welcome.

The older limitation above is now partially superseded. The repository has Python property-based fuzz coverage for high-risk structured inputs, including bitstream/IR parsing, NIR import, model-zoo .npz inputs, chip-spec JSON, optimizer evidence JSON, SCPN datastream JSON, Studio graph JSON, equation-to-MLIR lowering, HDL-source lowering, and transfer-checkpoint inputs.

Current release-blocking security checks:

pytest tests/test_fuzz_*.py tests/test_hypothesis_properties.py -q
python tools/supply_chain_audit.py --strict
bandit -r src/sc_neurocore/ -c pyproject.toml -q

Remaining hardening work:

• Add dedicated Rust cargo-fuzz targets for native parsers and PyO3 boundary adapters.
• Keep expanding Python Hypothesis coverage for malicious .nir, .npz, JSON, and pathological bitstream-length cases.
• Complete an external third-party security review before making audited security claims.
• No paid bug-bounty program is active yet. Until a scoped bounty is funded, coordinated disclosure through GitHub Security Advisories and the security email above is the official process.