Skip to content

Security: ResearchForumOnline/ZSEC

Security

SECURITY.md

Security Policy

ZSEC Auto Updates is a security tool, so the project keeps a strict boundary between advisory data and server actions.

Supported Scope

Primary support:

  • Ubuntu 22.04, 24.04, and compatible newer releases
  • AlmaLinux 8/9/10 compatible releases
  • Rocky Linux 8/9/10 compatible releases

Best-effort support:

  • Debian-like systems
  • Proxmox VE hosts
  • Linux Mint/Ubuntu-derived systems used as servers

Reporting a Security Issue

Please do not publish exploit details publicly before the maintainer has had a fair chance to review them.

Report non-sensitive issues through the repository issue tracker. For sensitive vulnerability details, email admin@talktoai.org with the subject ZSEC security report. Do not open a public issue containing exploit details or private data.

Useful report details:

  • affected ZSEC version or commit
  • operating system and version
  • exact command used
  • expected behavior
  • observed behavior
  • logs with secrets removed

Never include passwords, private keys, API keys, tokens, or full server credential files.

Security Design Promises

ZSEC must not:

  • run commands from the advisory feed
  • install packages named by the advisory feed
  • change firewall rules from the advisory feed
  • change SSH policy from the advisory feed
  • require AI, an API key, or a local model to run
  • silently switch from security updates to feature upgrades

The separate ZSEC Shield v0.1.2 prerelease is deterministic and no-AI, but its published native archives are unsigned. It must not be described as certified antivirus, real-time protection, guaranteed detection, or Store-approved software. Verify authenticated checksums, test in a safe environment, and keep operating-system protection enabled.

ZSEC may:

  • apply OS security updates through apt or dnf
  • create local advisory TODOs
  • back up SSH files before hardening
  • configure fail2ban for SSH
  • warn about public AI/dev-like listeners
  • apply conservative sysctl hardening

Owner Automation

The owner may use Codex/OpenAI to review public advisories, update documentation, rebuild the public feed, and push repository updates. That owner workflow is not part of the server runtime and must not create a public remote-control channel.

There aren't any published security advisories