CVE-2023-38470 is a reachable assertion (CWE-617) in Avahi's avahi_escape_label() function. A local process that asks avahi-daemon to resolve a crafted host name, for example with avahi-resolve -n, makes the daemon abort, and the device drops out of mDNS discovery. NVD scores it 5.5; commits 94cb648 and 20dec84 in v0.9-rc1 fix it.
Vulnerability summary
CVSS base score
5.5
Medium severity
EPSS score (FIRST)
0.3%
Probability of exploitation in the wild in the next 30 days
CISA KEV catalog
No
Not listed
Published
Vulnerability ID: CVE-2023-38470 · Source: NVD
Affected components
| Component | CPE | Version range |
|---|---|---|
| Avahi (avahi-daemon) | cpe:2.3:a:avahi:avahi:*:*:*:*:*:*:*:* | before 0.9 (NVD CPE); before 0.9-rc1 (CNA). Fixed upstream in v0.9-rc1 (commits 94cb648 and 20dec84); backported in Buildroot 0.8 patches 0005/0006, Yocto kirkstone/scarthgap CVE-2023-38470-1/-2, Debian 0.8-5+deb11u3 and 0.8-10+deb12u1 |
What does this mean for affected devices?
CVE-2023-38470 is one of five reachable assertions in Avahi published together on 2023-11-02 (CVE-2023-38469 to CVE-2023-38473). NVD describes it as "a reachable assertion" in the avahi_escape_label() function and lists Avahi before 0.9 as affected, plus Red Hat Enterprise Linux 8 and 9. Red Hat is the CNA (CVE record) and gives the affected range as versions before 0.9-rc1.
How is CVE-2023-38470 triggered?
The upstream report, Avahi issue 454, reproduces it with a host name lookup: avahi-resolve -n with a long crafted name made of short labels such as .}, .= and .?. avahi-daemon aborts on the assertion *ret_size > 0 in avahi_escape_label() (domain.c). The backtrace in the report shows the path: the D-Bus host name resolver request reaches avahi_s_host_name_resolver_prepare(), which validates the name through avahi_is_valid_fqdn() and avahi_normalize_name(). The title of the follow-up fix, "bail out when escaped labels can't fit into ret", names the underlying case: an escaped label that no longer fits the output buffer.
Avahi's D-Bus policy lets any local user call resolver methods, so no special group is needed. The oss-security post that announced the five CVEs ties the disable-user-service-publishing setting only to CVE-2023-38469 and CVE-2023-38472; a lookup publishes nothing, so that setting does not close this path.
How severe is CVE-2023-38470?
NVD's Primary score is CVSS v3.1 5.5 (CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H). Red Hat's CNA score is 6.2 (CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H), with no privileges required. NVD and Red Hat both list CWE-617 (reachable assertion). The impact is availability only: on a printer, camera, NAS or gateway, avahi-daemon stops and the device disappears from mDNS/DNS-SD discovery until the daemon restarts. It is not in the CISA KEV catalog as of 2026-11-21 (catalog version 2026.11.20, 1,790 entries), and FIRST EPSS gives it 0.0032, the 25.6th percentile, in the 2026-11-20 dataset.
Which firmware builds carry the fix?
Most embedded builds ship Avahi 0.8 with backports, so the version string alone does not tell you the status.
- Buildroot builds Avahi 0.8. Commit e728d350 of 2026-02-13 added
0005-ensure-each-label-is-at-least-one-byte-long.patch(upstream 94cb648) and0006-common-bail-out-when-escaped-labels-can-t-fit-into-ret.patch(upstream 20dec84) together with fixes for the other 2023 assertions and later Avahi CVEs, and lists CVE-2023-38470 inAVAHI_IGNORE_CVES. Buildroot 2026.02 contains the commit; images built from a tree without it carry the unpatched 0.8. - Yocto Project: poky kirkstone and scarthgap apply
CVE-2023-38470-1.patch(a backport of 94cb648) andCVE-2023-38470-2.patch(20dec84) to Avahi 0.8. - OpenWrt: the packages feed builds 0.9-rc5 on openwrt-24.10 and openwrt-25.12; openwrt-23.05 builds 0.8 (release 8) with no patch for this CVE as of 2026-11-21.
- Debian fixed bullseye in 0.8-5+deb11u3 with DLA-3990-1, bookworm in 0.8-10+deb12u1, trixie in 0.8-16 and unstable in 0.8-18; buster was postponed as a minor issue (Debian security tracker).
We found no device vendor or CISA advisory that names products for this CVE as of 2026-11-21.
How do you fix or mitigate CVE-2023-38470?
The fix is commit 94cb648 by Petr Menšík ("Ensure each label is at least one byte long"), merged on 2023-09-14, with the follow-up commit 20dec84 ("bail out when escaped labels can't fit into ret") of 2023-09-19; both are in tag v0.9-rc1. Rebuild with Avahi 0.9-rc1 or later, or apply both commits to 0.8, then confirm that the avahi-resolve -n call from issue 454 returns an error instead of killing the daemon. Where a rebuild is slow, make sure no untrusted local process (a container, a plugin runtime, a web UI user account) can reach the system D-Bus.
Our Avahi and mDNS/DNS-SD vulnerabilities guide covers version matching and hardening. The sibling assertions CVE-2023-38471 and CVE-2023-38472 need the same patch-list check, and the network-reachable CVE-2026-24401 is the Avahi flaw to rank above all of them.
Evidence
- Reachable assertion in avahi_escape_label(); NVD Primary CVSS v3.1 5.5 AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H; Red Hat Secondary 6.2 PR:N; CWE-617; CPE avahi before 0.9, RHEL 8.0 and 9.0; status Modified; published 2023-11-02; read 2026-11-21. — NVD
- State PUBLISHED; CNA Red Hat; title 'Reachable assertion in avahi_escape_label'; avahi affected from 0 before 0.9-rc1; CNA CVSS 3.1 6.2; checked 2026-11-21. — CVE.org
- avahi-resolve -n with a crafted name triggers 'domain.c:123: avahi_escape_label: Assertion *ret_size > 0 failed' through avahi_normalize_name, avahi_is_valid_fqdn and avahi_s_host_name_resolver_prepare from the D-Bus host name resolver; closed 2023-09-14. — Avahi issue 454
- 94cb648 'Ensure each label is at least one byte long' (Petr Menšík, committed 2023-09-14) and 20dec84 'bail out when escaped labels can't fit into ret' (committed 2023-09-19); both contained in v0.9-rc1 and not in v0.8 (GitHub compare). — Avahi commits
- Alan Coopersmith, 2023-10-06: CVE-2023-38470 is issue 454, fix 94cb648; disable-user-service-publishing named only for CVE-2023-38469 and CVE-2023-38472. — oss-security
- Buildroot e728d350 (2026-02-13) adds patches 0005 (94cb648, CVE: CVE-2023-38470) and 0006 (20dec84, CVE: CVE-2023-38470), AVAHI_IGNORE_CVES += CVE-2023-38470; contained in tag 2026.02. Yocto kirkstone and scarthgap avahi_0.8.bb list CVE-2023-38470-1.patch (backport of 94cb648) and CVE-2023-38470-2.patch (20dec84). OpenWrt packages openwrt-23.05 avahi 0.8 release 8 has no patch for it; openwrt-24.10 and openwrt-25.12 build 0.9_rc5; read 2026-11-21. — Buildroot / Yocto / OpenWrt
- bullseye fixed 0.8-5+deb11u3 (DLA-3990-1), bookworm 0.8-10+deb12u1, trixie 0.8-16, forky/sid 0.8-18; buster postponed (minor issue); read 2026-11-21. — Debian security tracker
- Not in the CISA KEV catalog version 2026.11.20 (1,790 entries), checked 2026-11-21. — CISA KEV
- EPSS 0.0032, percentile 0.25604, dataset 2026-11-20, read 2026-11-21. — FIRST EPSS

Related advisories
Sources
- NVD: CVE-2023-38470
- CVE.org record CVE-2023-38470 (CNA: Red Hat)
- Red Hat Bugzilla 2191690
- Avahi issue 454: reachable assertion in avahi_escape_label
- Avahi commit 94cb648 (fix)
- Avahi commit 20dec84 (follow-up fix)
- oss-security: CVEs assigned for reachable assertions in avahi (2023-10-06)
- Buildroot commit e728d350: package/avahi: patch various security vulnerabilities
- Yocto Project poky (kirkstone): avahi_0.8.bb
- OpenWrt packages (openwrt-23.05): libs/avahi/Makefile
- Debian security tracker: CVE-2023-38470
- Debian LTS Advisory DLA-3990-1: avahi security update
- CISA KEV
- FIRST EPSS
Frequently asked questions
Who can trigger CVE-2023-38470?
Any local user or process that can ask avahi-daemon over D-Bus to resolve a host name, for example with avahi-resolve -n. It is not reachable from the network, and setting disable-user-service-publishing does not block it because a lookup publishes nothing (per Avahi issue 454 and the oss-security announcement).
Which fix should a Buildroot or Yocto image carry?
Both upstream commits, 94cb648 and its follow-up 20dec84, which are in v0.9-rc1. Buildroot ships them as patches 0005 and 0006 on Avahi 0.8 since commit e728d350 (in 2026.02); Yocto kirkstone and scarthgap ship them as CVE-2023-38470-1.patch and CVE-2023-38470-2.patch (per the Buildroot and Yocto recipes).
How severe is CVE-2023-38470?
NVD scores it CVSS v3.1 5.5 (medium): local, availability impact only. Red Hat as CNA scores 6.2 because it rates no privileges required. The worst case on a device is avahi-daemon stopping until it is restarted (per NVD and CVE.org).
Is CVE-2023-38470 exploited in the wild?
It is not in the CISA KEV catalog as of 2026-11-21 (catalog version 2026.11.20), and FIRST EPSS gives it 0.0032, the 25.6th percentile, in the 2026-11-20 dataset (per CISA KEV and FIRST EPSS).
Questions about this page?