Skip to content

threat-model.md

Latest commit

 

History

History
89 lines (63 loc) · 2.16 KB

threat-model.md

File metadata and controls

89 lines (63 loc) · 2.16 KB

Threat Model

This document defines Ignite security goals, trust boundaries, and explicit non-goals.

Security Goals

Ignite aims to provide defense-in-depth for untrusted JS/TS execution.

Goal Mechanism
Reduce network exfiltration risk audit mode uses network-disabled container execution
Reduce filesystem tampering risk read-only root filesystem in audit mode
Limit privilege escalation dropped capabilities and no-new-privileges
Bound runaway workloads memory/cpu/time limits from service config
Preserve inspectability structured preflight and execution reporting

Trust Boundaries

Trusted:

  • host OS and kernel
  • Docker daemon
  • Ignite code and published artifacts

Untrusted:

  • service source code
  • service dependencies
  • runtime input payloads

Threat Classes

  • accidental unsafe code produced by LLM workflows
  • intentionally malicious user code
  • supply-chain risk from dependencies
  • abuse attempts via exposed HTTP API

What Audit Mode Helps With

  • network disablement by container flag
  • read-only root filesystem enforcement
  • capability stripping
  • security-event collection and reporting

Audit mode should be treated as the default for untrusted workloads.

Known Limits and Non-Goals

Ignite does not guarantee protection against:

  • Docker or kernel zero-days
  • side-channel attacks
  • malicious code behavior within allowed CPU/time budget
  • host compromise outside container boundary

Deployment Guidance

  • run Docker on dedicated hosts for untrusted workloads
  • keep Docker and host kernel patched
  • enable API auth when using HTTP server
  • place HTTP server behind trusted network boundary/proxy
  • set strict service resource limits
  • monitor logs and execution anomalies

Security Policy Files

When --audit is used, policy is loaded from one of:

  • ignite.policy.yaml
  • ignite.policy.yml
  • .ignite-policy.yaml
  • .ignite-policy.yml

Example:

security:
  network:
    enabled: false
  filesystem:
    readOnly: true
  process:
    allowSpawn: false
    allowedCommands: []

Responsible Disclosure

Please report vulnerabilities privately to the maintainer instead of opening public issues.