Skip to content

Security: BushidoCyb3r/salient

Security

SECURITY.md

Security Policy

Supported Versions

Salient is pre-1.0. Only the latest release gets fixes.

Version Supported
latest ✅
older ❌

Reporting a Vulnerability

Report privately via GitHub Security Advisories for this repository — do not open a public issue for a security bug.

Expect an initial response within a few days. Confirmed issues get a fix and a coordinated advisory; declined reports get an explanation.

Scope

Salient is a read-only, passive client: it queries an existing Elasticsearch/Security Onion deployment and writes only to the local filesystem (snapshots, reports, maps, operator registries, sanitized declared models, and optional analysis/tag sidecars). In scope for this policy:

  • Vulnerabilities in Salient's own code (the Go binary, the desktop GUI, the report/map renderers) — memory safety, injection into generated HTML/SVG/GraphML, unsafe handling of API keys or Elasticsearch responses, path traversal in snapshot/report file handling.

Out of scope:

  • Misconfiguration or vulnerabilities in your own Security Onion / Elasticsearch deployment.
  • Compromise of the Elasticsearch API key you supply to Salient — protect it like any other credential; Salient never transmits it anywhere but your configured Elasticsearch endpoint.
  • Optional analyze and desktop device-tagging calls to an operator-supplied model endpoint — that endpoint's security is your responsibility. Salient sends only the documented capped summary and requires explicit acknowledgement before remote network-data egress.

There aren't any published security advisories