CVE-2024-3094 is a backdoor (CWE-506) planted in the upstream XZ Utils 5.6.0 and 5.6.1 release tarballs: the build injects code into liblzma that can hijack sshd authentication on x86-64 glibc systems where sshd links libsystemd. NVD scores it 10.0. Remove 5.6.0 and 5.6.1; Debian and OpenWrt reverted to 5.4.x.
Vulnerability summary
CVSS base score
10
Critical severity
EPSS score (FIRST)
90.9%
Probability of exploitation in the wild in the next 30 days
CISA KEV catalog
No
Not listed
Published
Vulnerability ID: CVE-2024-3094 · Source: NVD
Affected components
| Component | CPE | Version range |
|---|---|---|
| XZ Utils (xz, liblzma) release tarballs | cpe:2.3:a:tukaani:xz:5.6.0:*:*:*:*:*:*:* | 5.6.0 and 5.6.1 (NVD, CVE record); clean releases from 2024-05-29, 5.6.2 onward (tukaani.org) |
| Debian xz-utils | pkg:deb/debian/xz-utils | testing, unstable and experimental 5.5.1alpha-0.1 up to and including 5.6.1-1; fixed in 5.6.1+really5.4.5-1 (DSA-5649-1); buster, bullseye, bookworm not affected (Debian security tracker) |
| OpenWrt tools/xz and packages feed utils/xz | — | 5.6.1 on the development branches from 2024-03-28 (tools) and 2024-03-20 (packages feed); reverted to 5.4.6 on 2024-03-29 (OpenWrt git history) |
What does this mean for affected devices?
CVE-2024-3094 is malicious code (CWE-506) in the upstream release tarballs of XZ Utils 5.6.0 and 5.6.1. NVD's description: "Through a series of complex obfuscations, the liblzma build process extracts a prebuilt object file from a disguised test file existing in the source code, which is then used to modify specific functions in the liblzma code." NVD's Primary score is 10.0, CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H. Red Hat, the CNA, gives the same 10.0 vector and rates it Critical. NVD's CPE configuration lists only tukaani:xz 5.6.0 and 5.6.1 (NVD, CVE record).
How does the backdoor work?
Andres Freund reported it on oss-security on 29 March 2024 after slow ssh logins on Debian sid. Part of the backdoor is "solely in the distributed tarballs": a modified build-to-host.m4 runs an obfuscated script at the end of configure. The script only injects code when building for x86-64 Linux (linux-gnu) with gcc and the GNU linker, as part of a Debian or RPM package build, and Freund judged it "can only work on glibc based systems". OpenSSH does not use liblzma directly, but several distributions patch it for systemd notification, and libsystemd depends on lzma; the injected code redirects RSA_public_decrypt in sshd. The XZ Utils backdoor page by maintainer Lasse Collin states that the 5.6.0 and 5.6.1 tarballs "were created and signed by Jia Tan", and that the xz-embedded and xz-java repositories "never had malicious or suspicious content". The Linux kernel's own XZ decompressor is XZ Embedded, per the kernel's XZ documentation.
Which distributions and build systems shipped it?
- Debian: DSA-5649-1 says no Debian stable versions are known to be affected; compromised packages were in testing, unstable and experimental from 5.5.1alpha-0.1 (uploaded 2024-02-01) up to and including 5.6.1-1, reverted to upstream 5.4.5 as 5.6.1+really5.4.5-1. The Debian security tracker marks buster, bullseye and bookworm not affected ("Vulnerable code not present").
- OpenWrt: on the main development branch, the host tool
tools/xzwas updated to 5.6.1 on 2024-03-28 and reverted to 5.4.6 on 2024-03-29. The packages feed'sutils/xzwas updated to 5.6.1 on 2024-03-20 and reverted to 5.4.6 on 2024-03-29. - Buildroot: the
package/xzhistory on master goes from 5.4.6 to 5.4.7 to 5.6.2; the 5.6.2 bump states "This version skips the 5.6.0 and 5.6.1 releases which had the backdoor." - Yocto Project / OpenEmbedded-Core: the xz recipe history on master moves from 5.4.6 directly to 5.6.2 in an August 2024 commit described as "the first post-backdoor release".
No device vendor PSIRT or CISA ICS advisory naming an affected device product was found as of 2026-11-19. For firmware, check the xz and liblzma versions in the SBOM and in the build host tools; our Yocto and Buildroot SBOM guide covers where both are recorded.
How do you fix CVE-2024-3094?
Remove 5.6.0 and 5.6.1 from images, toolchains and build hosts. Debian reverted to 5.4.5 code; OpenWrt to 5.4.6. Collin's page reports new, clean XZ Utils releases on 2024-05-29, and Buildroot and OpenEmbedded-Core moved to 5.6.2. Rebuild any image produced on a host that had the compromised packages.
Is CVE-2024-3094 exploited?
It is not in the CISA KEV catalog as of 2026-11-19 (catalog version 2026.11.18). CISA's ADP SSVC entry records exploitation as "none", automatable "yes" and technical impact "total". FIRST EPSS gives it 0.909, the 99.8th percentile, in the 2026-11-18 dataset (CISA KEV, FIRST EPSS). Our NVD backlog and Vulnrichment guide uses its EPSS query as an example. Because the backdoor targets sshd, see also CVE-2024-6387 in OpenSSH, and CVE-2018-25032 in zlib for another compression library in firmware.
Evidence
- NVD: status Modified; [email protected] Primary CVSS 3.1 10.0 AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H; Red Hat Secondary 10.0 same vector; CWE-506 (Red Hat); CPE tukaani:xz 5.6.0 and 5.6.1; published 2024-03-29; read from cached API response 2026-11-19. — NVD
- CVE record (CNA Red Hat, PUBLISHED 2024-03-29): title 'Xz: malicious code in distributed source'; xz 5.6.0 and 5.6.1 affected; Red Hat severity Critical, CVSS 3.1 10.0; CWE-506; RHEL 6-10 listed with default status unaffected; CISA ADP SSVC exploitation none, automatable yes, technical impact total. — CVE.org
- oss-security, Andres Freund, 2024-03-29: backdoor in upstream xz tarballs 5.6.0 and 5.6.1, partly solely in distributed tarballs (build-to-host.m4); injection conditions x86-64 linux-gnu, gcc and GNU ld, Debian or RPM package build; likely only works on glibc systems; openssh patched for systemd notification links libsystemd which depends on lzma; RSA_public_decrypt redirected into backdoor code. — Openwall oss-security
- tukaani.org XZ Utils backdoor page (last updated 2025-01-17): 5.6.0 and 5.6.1 release tarballs contain a backdoor, created and signed by Jia Tan; xz-embedded and xz-java never had malicious or suspicious content; new clean XZ Utils releases made 2024-05-29; 5.6.2 as cleanup step. — Tukaani project
- Debian DSA-5649-1 (2024-03-29): no Debian stable versions known to be affected; compromised packages in testing, unstable, experimental from 5.5.1alpha-0.1 (2024-02-01) up to and including 5.6.1-1; reverted to upstream 5.4.5 as 5.6.1+really5.4.5-1. Tracker: buster, bullseye, bookworm not affected, vulnerable code not present. — Debian
- OpenWrt: openwrt.git commit 714c91d1a63f 'tools/xz: update to 5.6.1' (authored 2024-03-28) reverted by d4b6b7644320 on 2024-03-29 back to PKG_VERSION 5.4.6 citing CVE-2024-3094; packages.git commit 1cef352d2884 'xz: update to 5.6.1' (2024-03-20) reverted by 9e46b47ca3ec on 2024-03-29 to 5.4.6. — OpenWrt git (GitHub mirror)
- Buildroot package/xz history on master: 5.4.6 (bd172e52e7f4, 2024-02-13), 5.4.7 (40240ac30afc), 5.6.2 (d1d77eb27476, 2024-06-12) whose message says 'This version skips the 5.6.0 and 5.6.1 releases which had the backdoor.' — Buildroot git (GitHub mirror)
- OpenEmbedded-Core meta/recipes-extended/xz history on master: 5.4.5 -> 5.4.6 (9cc6c809c154, 2024-02-16), 5.4.6 -> 5.6.2 (6446d82a533d, authored 2024-08-22) 'This is the first post-backdoor release.' — OpenEmbedded-Core git (GitHub mirror)
- Linux kernel Documentation/staging/xz.rst: 'The XZ decompressor in Linux is called XZ Embedded.' — Linux kernel source
- Not in CISA KEV catalog version 2026.11.18; EPSS 0.909, percentile 0.9981, dataset 2026-11-18, read 2026-11-19. — CISA KEV / FIRST EPSS

Related advisories
Sources
- NVD
- CVE.org record CVE-2024-3094 (CNA: Red Hat)
- Openwall oss-security: backdoor in upstream xz/liblzma leading to ssh server compromise (2024-03-29)
- Tukaani: XZ Utils backdoor
- Debian DSA-5649-1: xz-utils security update
- Debian security tracker: CVE-2024-3094
- OpenWrt: Revert "tools/xz: update to 5.6.1" (CVE-2024-3094)
- OpenWrt packages: Revert xz update to 5.6.1 (CVE-2024-3094)
- Buildroot: package/xz: bump version to 5.6.2
- OpenEmbedded-Core: xz: upgrade 5.4.6 -> 5.6.2
- Linux kernel documentation: XZ data compression in Linux
- CISA KEV
- FIRST EPSS
Frequently asked questions
Which xz versions contain the CVE-2024-3094 backdoor?
The XZ Utils 5.6.0 and 5.6.1 release tarballs, which were created and signed by Jia Tan. Part of the backdoor exists only in the distributed tarballs, not in the git repository's build files (per tukaani.org, NVD and the oss-security report).
Is the Linux kernel's XZ decompressor affected by CVE-2024-3094?
The kernel's XZ decompressor is XZ Embedded, and the XZ maintainer states the xz-embedded repository never had malicious or suspicious content. The backdoor sits in the XZ Utils liblzma build (per tukaani.org and the kernel's XZ documentation).
Did OpenWrt, Buildroot or Yocto ship xz 5.6.0 or 5.6.1?
OpenWrt's development branches moved to 5.6.1 in March 2024 and reverted to 5.4.6 on 2024-03-29. Buildroot and OpenEmbedded-Core went from 5.4.x to 5.6.2, skipping 5.6.0 and 5.6.1 (per the OpenWrt, Buildroot and OpenEmbedded-Core git histories).
Was Debian stable affected by CVE-2024-3094?
No. DSA-5649-1 states no Debian stable versions are known to be affected; testing, unstable and experimental carried 5.5.1alpha-0.1 up to 5.6.1-1 and were reverted to 5.4.5 code as 5.6.1+really5.4.5-1 (per Debian DSA-5649-1 and the Debian security tracker).
Is CVE-2024-3094 in the CISA KEV catalog?
No, as of 2026-11-19 (catalog version 2026.11.18). CISA's SSVC entry records exploitation as none. FIRST EPSS gives it 0.909, the 99.8th percentile, in the 2026-11-18 dataset (per CISA KEV, the CVE record and FIRST EPSS).
Questions about this page?