Which Linux platforms Pavois has actually proved

9 of 9 supported systems have been through a full clean-VM campaign against the rule base as it stands today. Every row is generated from those campaigns by the same check that gates a release: a system nobody proved cannot appear as proved.

9/9
proved on a clean VM
0
curated, never campaigned
30 d
freshness window
2026-09-21
matrix generated
PlatformStateLast campaignGrade before → afterPavois
Debian 12debian 12.15VERIFIED2026-09-210 days ago · workstationED509/542 (94%)dev
Debian 13debian 13.7VERIFIED2026-09-210 days ago · workstationED512/544 (94%)dev
Ubuntu 22.04ubuntu 22.04VERIFIED2026-09-210 days ago · workstationED488/526 (93%)dev
Ubuntu 24.04ubuntu 24.04VERIFIED2026-09-210 days ago · workstationED532/567 (94%)dev
Ubuntu 26.04ubuntu 26.04VERIFIED2026-09-210 days ago · workstationEE534/570 (94%)dev
RHEL 8 / Rocky 8 / AlmaLinux 8almalinux 8.10VERIFIED2026-09-210 days ago · workstationEE535/600 (89%)dev
RHEL 9 / Rocky 9 / AlmaLinux 9almalinux 9.8VERIFIED2026-09-210 days ago · workstationED580/626 (93%)dev
RHEL 10 / Rocky 10 / AlmaLinux 10almalinux 10.2VERIFIED2026-09-210 days ago · workstationEE549/605 (91%)dev
Fedorafedora 43VERIFIED2026-09-210 days ago · workstationEE569/622 (91%)dev

Freshness window: 30 days. Past it, a passing campaign turns STALE even when the rule base has not moved: the target’s own packages and kernel have, and no digest of ours can see that.

The Fedora corpus targets no release number: the norms reference records that no CIS benchmark exists for Fedora and that its mappings are inherited from RHEL 9. A verdict there covers the version actually measured, the one named under the label, not the distribution as a whole.

What each state means

VERIFIED

The last campaign passed, on this exact rule base, inside the freshness window.

FAILED

The last campaign ran and a required stage failed. An older green run does not override it.

STALE

It passed, but on a different rule base, or longer ago than the freshness window allows.

INCOMPLETE

A campaign exists and does not say what it proved: interrupted, or silent about which rule base it measured.

CURATED

The rule corpus exists, curated and statically validated. No clean-VM campaign has ever run. Not a weaker pass: a different claim.

UNKNOWN

Nothing to go on. Missing evidence never reads as a pass.

Why this page exists

For a hardening tool, release velocity is a poor maturity signal. A good one is what the campaigns have actually demonstrated, machine by machine, and what nobody has demonstrated yet. Both are published here.

A campaign is a clean VM created for it: scan the stock machine, plan, apply, reboot, a second pass re-planned from the current state, reboot, scan again, diff, then an evidence bundle that is verified. A container cannot stand in for it: 259 of the controls are meaningless without a kernel of your own.

The raw data behind this page is served as is, with machine identity removed: platform-matrix.json.