>> All posts

CVE-2022-26134: Atlassian Confluence OGNL injection, unauthenticated RCE, exploited as a zero-day

The short version: CVE-2022-26134 is an OGNL injection vulnerability in Atlassian Confluence Server and Data Center that lets an unauthenticated attacker execute code on the server. It was exploited as a zero-day before Atlassian shipped a fix, it scores 9.8, and it's in CISA's KEV catalog with one of the tightest remediation windows they've issued. If you run self-hosted Confluence, patch and verify. Steady hands, but move.

At a glance

FactDetail
Our severity takeHigh — in practice and on paper
CVSS v3.1 (NVD)9.8 — Critical · CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
EPSS~99.99% · 99.99th percentile (2026-07-17)
In CISA KEV?Yes — remediation was due 2022-06-06
Known exploited?Yes — exploited as a zero-day before the patch existed
Vulnerability typeCWE-917 expression-language (OGNL) injection → unauthenticated remote code execution
Requires authenticated session?No — pre-authentication
AffectedConfluence Server & Data Center: 1.3.0–7.4.16; 7.13.0–7.13.6; 7.14.0–7.14.2; 7.15.0–7.15.1; 7.16.0–7.16.3; 7.17.0–7.17.3; 7.18.0
Fixed in7.4.17 · 7.13.7 · 7.14.3 · 7.15.2 · 7.16.4 · 7.17.4 · 7.18.1 (and later)

What you need to know

Like its 2021 predecessor, this is an OGNL injection: a crafted expression in the request URL is evaluated by the server, yielding arbitrary command execution on the Confluence host. What sets it apart is the timeline.

  • It was a zero-day. Attackers were using it in the wild before a patch existed — Atlassian and researchers documented active exploitation as the advisory went out.
  • The affected range is enormous. Atlassian noted that all supported (and many unsupported) versions were affected, so "we're on an old build" is not a safe assumption.
  • It's unauthenticated and internet-reachable. No credential, no user interaction, no chain.

How serious we see it

High — the 9.8 matches the reality.

Unauthenticated remote code execution on an internet-facing collaboration server, confirmed exploited before a fix shipped, is textbook top-of-queue. The reassuring, bounded part is that the fixes are published and exposure is quick to confirm. The catch is the zero-day window: any instance that was internet-facing before you patched should be treated as possibly touched until you've checked.

Recommendations

Straight from Atlassian's advisory and CISA:

  1. Patch now. Upgrade to 7.4.17 / 7.13.7 / 7.14.3 / 7.15.2 / 7.16.4 / 7.17.4 / 7.18.1 or later.
  2. Assume nothing given the zero-day window — hunt. Review logs for suspicious requests and check the host for web shells and other implants.
  3. If you find indicators, respond fully. Rebuild the host and rotate all credentials, tokens, and secrets it could reach.
  4. Harden exposure. Restrict internet reachability of Confluence; put it behind SSO/VPN where you can.
  5. Confirm your exposure first. Verify whether you run affected self-hosted Confluence (Confluence Cloud is not affected).

How BreachRisk sees it

BreachRisk answers the first question — do we have an exposed, affected Confluence instance? — from the outside in. Starting from little more than your domain, it discovers internet-facing Confluence, fingerprints the product and version, and flags exposure tied to KEV-listed, actively exploited vulnerabilities like this one, surfaced at the top because it's in KEV. We detect and flag the vulnerable version; we don't exploit it.

When a Confluence zero-day breaks, the outside-in view means the exposed server is already on your map — you go straight to patch-and-verify instead of starting from an inventory.

References

See your cyber risk, proven.