Skip to content

Security: AnimusHQ/mllaboratory

Security

SECURITY.md

Security Policy

Animus Datalab is an enterprise ML laboratory that treats reproducibility, auditability, and security as platform properties. Please report suspected vulnerabilities privately.

Reporting a vulnerability

Do not open a public GitHub issue for security-sensitive reports.

Send reports to:

rewanderer@proton.me

Please include, when possible:

  • affected component or path;
  • affected version, branch, or commit;
  • reproduction steps;
  • observed impact;
  • logs or traces with secrets removed;
  • whether the issue affects local development, Kubernetes deployment, Control Plane services, Data Plane execution, authentication, RBAC, audit, object storage, or supply-chain controls.

Security-sensitive areas

The following areas are especially sensitive:

  • Gateway authentication and session handling;
  • project-scoped RBAC and role-gated API surfaces;
  • internal-only Control Plane / Data Plane channels;
  • execution-bound run tokens and workload credentials;
  • secret injection and secret-provider integrations;
  • audit event integrity and append-only behavior;
  • immutable dataset, run, policy, and execution-plan records;
  • object-store references, content hashes, and artifact mediation;
  • Helm values, image pinning, and signed-image policy controls;
  • boundaries between Control Plane governance code and Data Plane runtime execution internals.

Security posture

The expected platform posture is deny-by-default:

  • unauthenticated or insufficiently scoped requests should fail closed;
  • mutating operations should require editor-grade or stronger authorization;
  • admin-oriented audit, lineage, and quality surfaces should remain restricted;
  • internal CP/DP APIs must not be exposed as public API surfaces;
  • secrets must not be logged, returned to UI clients, or persisted as general application state;
  • workload state is not authoritative until reconciled into PostgreSQL-backed Control Plane state.

Supported versions

Until the project publishes versioned releases, security assessment should target the current main branch and any tagged releases once available.

Disclosure

Animus will prioritize triage for reports that include a clear impact path. Coordinated disclosure timing should be agreed before public discussion.

There aren’t any published security advisories