R-AI-OS runs on your machine with elevated trust: it spawns AI agent subprocesses, holds a hash-chained audit ledger, and mediates tool access through its own security kernel (UMAI, policy manager, egress filter). Taking vulnerability reports seriously here is not optional.
Only the latest released version (main/master at the current tag, see
CHANGELOG.md) receives security fixes. There is no LTS branch
at this stage of the project.
| Version | Supported |
|---|---|
| Latest tagged release | ✅ |
| Older releases | ❌ |
Do not open a public GitHub issue for security vulnerabilities.
Preferred: use GitHub's private vulnerability reporting for this repository (repo → Security tab → Report a vulnerability). This creates a private advisory visible only to the maintainer until a fix ships.
If that's not available to you, email goktugturhan74@gmail.com with:
- A clear description of the vulnerability and its impact.
- Steps to reproduce (a minimal PoC is ideal).
- The affected version/commit.
You should get an acknowledgment within a few days. This is a small, single-maintainer project — please be patient, but a real report will always get a real response.
In scope:
- The
raios/aiosdbinaries and all crates undercrates/. - The MCP tool surface, the
/a2aHTTP endpoints, and the daemon's WebSocket dispatch. - The security kernel: filesystem jail, policy manager, verify-chain (audit ledger), egress filter, UMAI, secret leasing, rate limiting, quarantine.
Out of scope:
- Vulnerabilities that require an attacker to already have arbitrary code
execution as your local user (this tool trusts the local user by design —
see the Security Kernel section in
README.md). - Denial-of-service reports against your own local daemon instance.
- Findings from automated scanners without a demonstrated, concrete impact
(raw
raios securitypattern-scanner output on this repo's own codebase is not itself a report — see the caveat about that scanner's false-positive rate in this repo's operating notes).
We follow coordinated disclosure: please give us a reasonable window to ship
a fix before any public write-up. We'll credit reporters (unless you prefer
otherwise) in the CHANGELOG.md entry for the fix.