Skip to content

Security: EdgeFirstAI/hal

Security

SECURITY.md

Security Policy

Supported Versions

We actively support the following versions of EdgeFirst HAL with security updates:

Version Support Status
0.x βœ… Full support
< 0.1 ❌ Not yet released

Note: Version support policy will be updated once we reach stable 1.0 release.

Reporting a Vulnerability

Au-Zone Technologies takes security seriously across our entire EdgeFirst ecosystem, including the Hardware Abstraction Layer.

How to Report

Email: support@au-zone.com
Subject: "Security Vulnerability - HAL"

For GitHub Users: You can also report vulnerabilities privately through GitHub Security Advisories

Please include:

  • Vulnerability description - Clear explanation of the security issue
  • Steps to reproduce - Detailed reproduction steps
  • Affected versions - Which versions are impacted
  • Potential impact - Assessment of severity and risk
  • Suggested fixes - If you have ideas for remediation (optional)
  • Proof of concept - Code or configuration demonstrating the issue (if applicable)

What to Expect

  1. Acknowledgment within 48 hours of your report
  2. Initial assessment within 7 business days
  3. Fix timeline based on severity:
    • Critical: 7 days
    • High: 30 days
    • Medium: Next minor release
    • Low: Next major release

Response Process

Once we receive your report:

  1. We'll confirm receipt and begin investigation
  2. We'll assess severity using CVSS scoring
  3. We'll develop and test a fix
  4. We'll coordinate disclosure timing with you
  5. We'll release the fix and publish an advisory
  6. We'll update affected users through multiple channels

Responsible Disclosure

We ask that you:

  • Allow reasonable time for us to fix vulnerabilities before public disclosure
  • Avoid public disclosure until we've released a patch and advisory
  • Not exploit vulnerabilities for any purpose other than verification
  • Keep findings confidential until coordinated disclosure
  • Act in good faith toward our users and the security community

We commit to:

  • Acknowledge your report promptly
  • Keep you informed throughout the remediation process
  • Credit you in advisories (unless you prefer to remain anonymous)
  • Work collaboratively to understand and resolve the issue

Security Update Process

Security updates are released through:

  1. GitHub Security Advisories - Published on our repository
  2. Release Notes - Documented in CHANGELOG.md
  3. crates.io Updates - New versions with security fixes
  4. PyPI Updates - Python package updates
  5. Email Notifications - For users who subscribe to security alerts

Subscribing to Security Updates

Recognition

With your permission, we'll credit you in:

  • Security advisories - Public acknowledgment of your contribution
  • Release notes - Recognition in version release documentation
  • Annual security report - Listed as a security contributor
  • Hall of fame - On our website (if you prefer)

If you prefer to remain anonymous, we'll respect that choice.

Security Best Practices

When using EdgeFirst HAL in production:

Input Validation

  • Validate all external inputs before processing
  • Sanitize file paths and user-provided data
  • Check tensor dimensions before allocation
  • Verify image formats match expectations

Memory Safety

  • The HAL is written in Rust, providing memory safety guarantees
  • Use safe APIs and avoid unsafe blocks when possible
  • Be cautious with FFI boundaries (G2D, OpenGL)
  • Monitor memory allocation limits

Dependency Management

  • Keep dependencies updated - Regularly update to latest versions
  • Review security advisories - Check for known vulnerabilities
  • Use cargo audit - Scan for security issues in dependencies
cargo install cargo-audit
cargo audit

Hardware Acceleration

  • Validate hardware capabilities before use
  • Handle fallback paths securely
  • Limit resource consumption on shared hardware
  • Isolate untrusted inputs from accelerators

Python Bindings

  • Validate numpy arrays before conversion
  • Check data types match expectations
  • Handle exceptions properly
  • Limit memory exposure across FFI boundary

Known Security Considerations

DMA-Heap Memory

  • DMA-heap buffers can be shared between processes
  • File descriptors can be passed via Unix sockets
  • Ensure proper access control on shared buffers
  • Consider security implications in multi-tenant environments

Hardware Accelerators

  • G2D and OpenGL access may require elevated privileges
  • Shared GPU memory can be a side-channel
  • Validate all parameters passed to hardware APIs
  • Monitor for resource exhaustion attacks

Image Processing

  • Malformed images can cause decoder crashes
  • Large images can exhaust memory
  • Implement size limits for untrusted inputs
  • Use safe image decoding libraries

Additional Security Services

For production deployments requiring enhanced security:

Enterprise Security Offerings

Au-Zone Technologies provides:

  • Security Audits - Comprehensive code and architecture reviews
  • Penetration Testing - Third-party security assessments
  • Compliance Certification - Help meeting regulatory requirements
  • Priority Security Patches - Expedited fixes for enterprise customers
  • Custom Security Hardening - Tailored security enhancements
  • Dedicated Security Support - Direct access to security team

Professional Services

  • Secure Integration - Help integrating HAL securely into your stack
  • Threat Modeling - Assess risks for your specific use case
  • Security Training - Educate your team on secure usage
  • Incident Response - Support during security incidents

πŸ“§ Contact: support@au-zone.com
🌐 Learn more: au-zone.com/security

Security Disclosure History

We maintain transparency about security issues:

No security vulnerabilities have been disclosed to date.


Last Updated: November 2025

Thank you for helping keep EdgeFirst HAL and our users safe! πŸ”’

There aren't any published security advisories