Skip to content

Security: openairymax/agentrt-linux

Security

SECURITY.md

Security Policy

Security authority: docs/AirymaxOS/50-engineering-standards/09-ssot-registry.md (OS-SEC-* rules) Architecture threat model: docs/AirymaxOS/10-architecture/07-threat-model.md

Copyright (c) 2025-2026 SPHARX Ltd. All Rights Reserved.


1. Supported Versions

Version Supported Status
1.0.1 Yes Active development (Pilot)
0.1.1 Yes Design baseline (Experimental)
< 0.1.1 No End of life

2. Reporting a Vulnerability

2.1 Private Disclosure

Do NOT file public GitHub issues for security vulnerabilities.

Report vulnerabilities privately:

  1. Email: security@spharx.com
  2. Subject: [SECURITY] agentrt-linux — <brief description>
  3. PGP: Request public key from security@spharx.com

Include in your report:

  • Description of the vulnerability
  • Affected component (kernel / services / security / memory / cognition / cloudnative / system / tests-linux)
  • Steps to reproduce
  • Impact assessment
  • Suggested fix (if any)

2.2 Response Timeline

Step Timeline
Acknowledgment Within 48 hours
Initial assessment Within 7 days
Fix or mitigation Within 30 days (severity-dependent)
Public disclosure After fix release + 14-day grace period

2.3 Severity Classification

We use CVSS v3.1:

Severity CVSS Range Example
Critical 9.0–10.0 Kernel privilege escalation, RCE
High 7.0–8.9 Memory corruption, sandbox escape
Medium 4.0–6.9 Information leak, DoS
Low 0.1–3.9 Minor info disclosure

3. Security Design Principles

3.1 OS-IRON-001: User-space ABI Never Breaks

The include/uapi/airymax/ interface is permanently stable. Changes that break userspace are rejected.

3.2 Capability-Based Security (OS-SEC-*)

agentrt-linux adopts seL4-style capability-based security:

  • Every kernel resource access requires a capability
  • Capabilities are unforgeable tokens
  • No ambient authority — capabilities must be explicitly passed

3.3 [SC] Shared Contract Layer Integrity

The 10 core header files in kernel/include/uapi/linux/airymax/ are the single physical source of truth for cross-subsystem contracts. Any change requires:

  1. Dual CI validation (agentrt-linux + agentrt)
  2. L1 subsystem maintainer review
  3. L3 top maintainer (SPHARX Engineering) final approval

3.4 Security as Cross-Cutting Concern

Security is not a separate data flow — it is a cross-cutting concern that permeates all 4 major data flows (IPC, scheduling, memory, cognition). Every subsystem must implement the relevant OS-SEC-* rules.

4. Security CI Gates

The CI pipeline enforces:

  • checkpatch.pl: Style and common bug detection
  • sparse: Static analysis for kernel code
  • Coccinelle: Pattern-based bug detection
  • clang-analyzer: Additional static analysis
  • Code coverage: Minimum coverage thresholds per subsystem
  • SSoT validation: Rule ID consistency

5. Disclosure Policy

  • We follow Coordinated Disclosure.
  • Reporters are credited (unless they prefer anonymity).
  • We request a 90-day embargo before public disclosure.
  • Extensions may be granted for complex fixes.

There aren't any published security advisories