Skip to content

Security: Aymwvn/Detect-AI

Security

SECURITY.md

Security Policy

DetectAI processes security telemetry — alerts, logs, process/network data — which is itself sensitive and, per the project's own threat model (see docs/ARCHITECTURE.md, section 8), potentially adversarial. Security issues in this project are taken seriously.

Reporting a vulnerability

Please do not open a public GitHub issue for security vulnerabilities.

Instead, report privately by opening a GitHub Security Advisory on this repository (Security tab → "Report a vulnerability"), or by contacting the maintainer directly.

Please include:

  • A description of the vulnerability and its potential impact
  • Steps to reproduce (or a proof-of-concept)
  • Affected version/commit

You should expect an initial response within a few days. This is a student-maintained open-source project, not a funded security team, so response times will vary — but every report is read.

Scope

Security-relevant areas of particular interest:

  • Prompt injection: any way untrusted alert content (command_line, description, raw_event, etc.) could cause the AI layer to treat data as instructions, or bypass the evidence-validation boundary (architecture doc §15).
  • Authentication / authorization: RBAC bypass, privilege escalation, token handling.
  • Secret handling: credentials or API keys appearing in logs, error messages, API responses, or committed config.
  • Input validation: malformed or oversized connector payloads causing crashes, resource exhaustion, or injection into the database layer.
  • Container/deployment hardening: issues with the Docker Compose setup, non-root enforcement, network segmentation.

Supported versions

This project is pre-1.0 and under active phased development (see the roadmap in docs/ARCHITECTURE.md). Only the latest commit on main is currently supported — there is no long-term-support branch yet.

Out of scope

  • Vulnerabilities in third-party SIEM/EDR products that DetectAI connects to — report those to the respective vendor.
  • Vulnerabilities requiring an already-compromised deployment (e.g. an attacker with direct database access) are documented as accepted risk in the threat model rather than tracked as bugs, unless they reveal a missing control DetectAI itself should have had.

There aren't any published security advisories