Skip to content

Security: 855princekumar/yawcam-ai-dockerized

Security

SECURITY.md

Security Policy

This repository provides containerization, orchestration, and edge deployment tooling around the upstream Yawcam-AI project. While this repo does not include compiled model binaries or proprietary code from upstream, operational security and responsible deployment practices still apply.

Supported Versions

The following build paths are intended for regular security awareness and maintenance:

Variant Support Status
CPU-only build (Ubuntu base) ✔ Supported
GPU-enabled CUDA build ✔ Supported
Alpine GLIBC hybrid build ⚠ Experimental
Distroless / minimal images ⚠ Future consideration

Security fixes and operational updates will be applied to supported variants only.


Reporting Vulnerabilities

If you discover:

  • Container escape vectors
  • Misconfigured permissions
  • Privilege elevation risks
  • Network exposure issues
  • Credential persistence weaknesses

Please report privately rather than as a public issue.

Private Reporting Contact:
devprincekumar (GitHub profile DM) or Email

A response will typically be provided within 72 hours.


Responsible Disclosure

Please do not:

  • Publish exploit details before we triage and confirm
  • Commit PoCs directly to the repo
  • Submit issues demonstrating access to your private environments

We will:

✔ acknowledge researcher credit
✔ document fixes
✔ ship updated Dockerfiles / compose files where needed


Security Scope Clarification

This project does NOT modify or publish Yawcam-AI internal logic. It packages and orchestrates runtime execution.

⚠ Therefore, vulnerabilities affecting:

  • Java runtime
  • FFmpeg / CVE risks
  • YOLO inference frameworks
  • Spring Boot or embedded DB

…must also be disclosed upstream to the original project:

https://github.com/yawcam/yawcam-ai

We recommend reporting directly to both maintainers when in doubt.


Secure Usage Guidelines

To reduce attack surface:

✔ Prefer volume-mapped persistent storage vs container-internal storage
✔ Run as non-root where possible (future hardening planned)
✔ If exposing UI remotely, front it behind reverse proxy authentication
✔ Disable or restrict event-based script execution hooks if unneeded
✔ On GPU systems, ensure NVIDIA runtime is patched


Long-term Hardening Roadmap

  • Non-root container builds
  • Distroless minimal runtime
  • Optional reverse-proxy authentication profile
  • Signature validation for downloaded model files

Security contributions are highly welcome.

There aren't any published security advisories