MiOS is one system built two ways at once: an immutable, bootc/OCI-shaped Fedora workstation (the whole OS is a single container image you bootc upgrade like a git pull and bootc rollback like a Ctrl-Z) that is also a local, self-hosted agentic AI operating system — a full inference + agent stack running on the operator's own hardware behind one OpenAI-compatible endpoint (MIOS_AI_ENDPOINT).

Both halves raise the security stakes, which is why MiOS treats hardening as a first-class, defense-in-depth layer rather than an afterthought:

• The image-mode half means the root filesystem is content-addressed and tamper-evident, so verification has to extend from the OCI layer all the way down to per-file integrity (MOK-signed modules, kernel-lockdown integrity, fapolicyd, USBGuard).
• The agentic half means the machine runs code on the operator's behalf and reasons about itself, so the agent plane — the agent-pipe/MiOS-Hermes orchestrator, the mios-llm-light inference lane (:11450, also embeddings) and gated heavy GPU lanes, and the PostgreSQL+pgvector memory — must be unprivileged, network-fenced, and sandboxed by default (Architectural Law 6, UNPRIVILEGED-QUADLETS).

This file is the top-level entry point for the security topic: a short pointer that tells you where the enforced posture lives, where the audit record lives, and how to report a vulnerability privately. It deliberately carries no hardening configuration itself — the canonical guide below is the single source of truth, and every control there cites the exact file that enforces it.

The full security posture (kargs, SELinux, fapolicyd, USBGuard, CrowdSec sovereign-mode IPS, kernel-lockdown integrity, MOK signing, kargs hardening) lives in the canonical FHS doc location:

• usr/share/doc/mios/guides/security.md

That guide is the defense-in-depth map: each control is tied to the exact file that enforces it (usr/lib/bootc/kargs.d/*.toml, the SELinux modules under usr/share/selinux/packages/mios/, the 33-firewall/37-selinux automation steps, and the mios-hardening/mios-firewall-init units), with the admin override for each.

Audit reports, including the AUDIT-FINDINGS-YYYYMMDD.md series produced by the read-only audit prompt at usr/share/mios/ai/audit-prompt.md, ship under usr/share/doc/mios/audits/.

These are a historical record of point-in-time reviews — they are kept as the audit trail and are not edited to track the current state. The live posture is always the canonical guide above.

For private disclosure of a vulnerability in 'MiOS', open a private security advisory at https://github.com/mios-dev/MiOS/security/advisories/new or contact the project maintainers via the channels listed in the GitHub repository's security tab. Do not file a public issue for security reports.