ASI:BUILD is research software in early alpha. We provide security fixes on a best-effort basis:
| Version | Supported |
|---|---|
main (latest) |
✅ Active development |
v0.1.0-alpha |
✅ Best-effort |
| Older | ❌ |
ASI:BUILD is a local Python research framework with no network-facing server components by default. The main security concerns are:
- Cypher/query injection —
graph_intelligenceandknowledge_graphmodules that accept user-provided queries - Dependency vulnerabilities — transitive dependencies (NumPy, NetworkX, PyTorch, etc.)
- Deserialization — pickle-based model loading, particularly in
optimizationandfederatedmodules - Credential handling —
configs/files that accept API keys or cloud credentials - MCP server (
servers/) — if exposed over the network, the MCP surface should be audited
Out of scope (by design):
- The
archive/directory — experimental v1 code, not maintained - Module stubs that do not execute real code
Please do not open a public GitHub issue for security vulnerabilities.
Report security issues via one of:
- GitHub private vulnerability disclosure: Use the "Report a vulnerability" button on the Security tab
- Email: Contact the maintainer via GitHub profile
- A description of the vulnerability and its impact
- Steps to reproduce (minimal code example preferred)
- Affected module(s) and version(s)
- Any proposed fix, if you have one
- Acknowledgment within 72 hours
- Initial assessment within 1 week
- Fix or advisory within 30 days for confirmed critical issues
We follow responsible disclosure: we will coordinate with you before any public disclosure and give credit in the security advisory.
The graph_intelligence module builds Cypher queries from user-provided strings. Always use parameterized queries rather than string interpolation with user input.
The optimization module can load PyTorch model checkpoints. Only load checkpoints from trusted sources. We plan to migrate to safetensors format in v0.3.0.
The federated module implements Byzantine-tolerant aggregation, but the network transport layer is stub-level in the current alpha. Do not use the federated module in a real networked deployment without a security review.
We run pip-audit as part of CI. To check your local install:
pip install pip-audit
pip-auditTo report a vulnerability in a dependency (not in our code), please report it to the upstream project. We will update the dependency when a fix is available.
Last updated: 2026-04-11