CVE-2026-16118: xdgmime: heap overflow in magic file parsing

HIGHCISA: TRACK*
Published July 17, 2026
CISO Take

CVE-2026-16118 is a heap-based buffer overflow in xdgmime's MIME magic file parser, triggered when an application performing MIME detection (for example via g_content_type_guess()) loads an attacker-planted magic file from a user-writable XDG data directory. It is a general-purpose library flaw and has no AI-specific attack surface, but it ships in glib2, webkitgtk and Red Hat AI base images (rhai/*, rhaiis/vllm-*), so AI containers built on RHEL 9 inherit it. Risk is moderate: CVSS 7.1 with local access and user interaction required, an EPSS of 0.28% (82nd percentile), no public exploit, no Nuclei template, no KEV listing, and CISA SSVC set to TRACK*. An attacker needs a prior foothold to write into the XDG data path, so this is a post-compromise memory-corruption primitive rather than an initial-access vector. Apply the Red Hat errata (RHSA-2026:64799 and the related advisories), rebuild AI images from patched base layers, and restrict write access to XDG data directories in shared or multi-tenant environments.

Sources:NVDEPSSRed Hat Security AdvisoriesATLAS

What is the risk?

Medium overall. Exploitability is limited by local access, the need for user interaction, and the need to write into $XDG_DATA_HOME/mime/magic. Impact is high for integrity and availability (a 2-byte out-of-bounds heap write causing a crash or memory corruption), with no confidentiality impact in the CVSS vector. EPSS is low (0.00277), there is no known exploit, and the CVE is not in KEV. Exposure for AI/ML systems is indirect: inference containers (vLLM on RHEL AI), notebook hosts and desktop AI tooling that link glib2 or webkitgtk are affected only if they parse MIME data from user-writable locations.

How does the attack unfold?

Initial Foothold
Attacker gains low-privilege local access or write ability to the user's XDG data directory on an AI host or container.
AML.T0010.001
Payload Planting
Attacker writes a crafted mime/magic file with a malformed entry under $XDG_DATA_HOME.
Exploitation
A MIME-detecting application such as one calling g_content_type_guess() parses the file and performs a 2-byte out-of-bounds heap write.
AML.T0011
Impact
The process crashes or its heap is corrupted, disrupting AI tooling and possibly enabling further code execution.

What systems are affected?

PackageEcosystemVulnerable RangePatched
vLLMpip—No patch
96.9K93 dependentsPushed 3d ago26% patched~37d to patchFull package profile →
vLLMpip—No patch
96.9K93 dependentsPushed 3d ago26% patched~37d to patchFull package profile →
cert-manager/cert-manager-istio-csr-rhel9——No patch
cert-manager/cert-manager-operator-rhel9——No patch
cert-manager/cert-manager-trust-manager-rhel9——No patch
cert-manager/jetstack-cert-manager-acmesolver-rhel9——No patch
cert-manager/jetstack-cert-manager-rhel9——No patch
glib2——No patch
rhai/base-image-cpu-rhel9——No patch
rhai/base-image-cuda-rhel9——No patch
rhai/base-image-rocm-rhel9——No patch
rhai/base-image-spyre-rhel9——No patch
rhai/base-image-tpu-rhel9——No patch
rhaiis/model-opt-cuda-rhel9——No patch
rhui5/cds-kubernetes-rhel9——No patch
rhui5/cds-rhel9——No patch
rhui5/haproxy-rhel9——No patch
rhui5/installer-rhel9——No patch
rhui5/rhua-rhel9——No patch
webkit2gtk3——No patch
webkitgtk4——No patch
xdgmime——No patch

How severe is it?

CVSS 3.1
7.1 / 10
EPSS
0.3%
chance of exploitation in 30 days
Higher than 20% of all CVEs
Exploitation Status
Exploit Available
Exploitation: MEDIUM
Sophistication
Advanced
Exploitation Confidence
medium
○CISA SSVC: Public PoC
Composite signal derived from CISA KEV, VulnCheck KEV, CISA SSVC, EPSS, Metasploit, Exploit-DB, trickest/cve, Nuclei templates, and inthewild.io exploitation reports.

What is the attack surface?

AVACPRUISCIA
AVLocal
ACLow
PRNone
UIRequired
SUnchanged
CNone
IHigh
AHigh

What should I do?

1 step
  1. 1) Install the fixed glib2, webkit2gtk3 and webkitgtk4 packages and the updated container images listed in the Red Hat errata (RHSA-2026:64799, 64800, 66451, 67956, 70636, 71403-71405, 72394, 72395). 2) Rebuild and redeploy AI/ML containers (rhai/base-image-*, rhaiis/vllm-*) from the patched base images and scan registries for stale layers. 3) Workaround: make $XDG_DATA_HOME/mime and /usr/share/mime read-only or root-owned for service accounts, and run inference workloads with read-only root filesystems and ephemeral home directories. 4) Detection: monitor for unexpected creation or modification of files under */mime/magic, and for crashes (SIGSEGV, SIGABRT, heap-corruption messages) in processes using glib content-type guessing.

What does CISA's SSVC say?

DecisionTrack*
Exploitationpoc
AutomatableNo
Technical Impactpartial

Source: CISA Vulnrichment (SSVC v2.0). Decision based on the CISA Coordinator decision tree.

How is it classified?

DoSCode ExecutionInferenceFrameworkAML.T0010.001AML.T0011

Which compliance frameworks are affected?

This CVE is relevant to:

ISO 42001
A.6.2.6-AI system operation and monitoring
NIST AI RMF
MANAGE 2.3-Respond to and manage previously unknown risks
OWASP LLM Top 10
LLM03-Supply Chain

Frequently Asked Questions

What is CVE-2026-16118?

CVE-2026-16118 is a heap-based buffer overflow in xdgmime's MIME magic file parser, triggered when an application performing MIME detection (for example via g_content_type_guess()) loads an attacker-planted magic file from a user-writable XDG data directory. It is a general-purpose library flaw and has no AI-specific attack surface, but it ships in glib2, webkitgtk and Red Hat AI base images (rhai/*, rhaiis/vllm-*), so AI containers built on RHEL 9 inherit it. Risk is moderate: CVSS 7.1 with local access and user interaction required, an EPSS of 0.28% (82nd percentile), no public exploit, no Nuclei template, no KEV listing, and CISA SSVC set to TRACK*. An attacker needs a prior foothold to write into the XDG data path, so this is a post-compromise memory-corruption primitive rather than an initial-access vector. Apply the Red Hat errata (RHSA-2026:64799 and the related advisories), rebuild AI images from patched base layers, and restrict write access to XDG data directories in shared or multi-tenant environments.

Is CVE-2026-16118 actively exploited?

No confirmed active exploitation of CVE-2026-16118 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-16118?

1) Install the fixed glib2, webkit2gtk3 and webkitgtk4 packages and the updated container images listed in the Red Hat errata (RHSA-2026:64799, 64800, 66451, 67956, 70636, 71403-71405, 72394, 72395). 2) Rebuild and redeploy AI/ML containers (rhai/base-image-*, rhaiis/vllm-*) from the patched base images and scan registries for stale layers. 3) Workaround: make $XDG_DATA_HOME/mime and /usr/share/mime read-only or root-owned for service accounts, and run inference workloads with read-only root filesystems and ephemeral home directories. 4) Detection: monitor for unexpected creation or modification of files under */mime/magic, and for crashes (SIGSEGV, SIGABRT, heap-corruption messages) in processes using glib content-type guessing.

What systems are affected by CVE-2026-16118?

This vulnerability affects the following AI/ML architecture patterns: model serving, containerized inference, notebook environments, developer workstations.

What is the CVSS score for CVE-2026-16118?

CVE-2026-16118 has a CVSS v3.1 base score of 7.1 (HIGH). The EPSS exploitation probability is 0.28%.

What is the AI security impact?

Affected AI Architectures

model servingcontainerized inferencenotebook environmentsdeveloper workstations

MITRE ATLAS Techniques

AML.T0010.001AI Software
AML.T0011User Execution

Compliance Controls Affected

ISO 42001: A.6.2.6
NIST AI RMF: MANAGE 2.3
OWASP LLM Top 10: LLM03

What are the technical details?

Original Advisory

A flaw was found in xdgmime. A heap-based buffer overflow can be triggered in _xdg_mime_magic_parse_magic_line() in the xdgmimemagic.c file on little-endian systems when an attacker-controlled MIME magic file in a user-writable XDG data location (e.g., in the $XDG_DATA_HOME/mime/magic path) is parsed by an application performing MIME type detection (e.g., via g_content_type_guess()). When performing byte-swap, incorrect pointer arithmetic on the write side causes an out-of-bounds write of 2 bytes, resulting in an application crash or memory corruption.

Exploitation Scenario

An attacker with limited local access to a shared GPU or notebook host (for example through a compromised low-privilege account or a malicious uploaded artifact that writes to the user's home) plants a crafted $XDG_DATA_HOME/mime/magic file with a malformed byte-swap entry. When a user or service later opens a file through an application that calls g_content_type_guess(), the parser performs a 2-byte out-of-bounds heap write on little-endian hardware. The result is a process crash that disrupts a notebook or tooling session, or heap corruption that a more capable attacker could try to turn into code execution in the user's context.

Weaknesses (CWE)

CWE-122 — Heap-based Buffer Overflow: The product writes past the end of a buffer that was allocated on the heap.

  • Pre-design: Use a language or compiler that performs automatic bounds checking.
  • [Architecture and Design] Use an abstraction library to abstract away risky APIs. Not a complete solution.

Source: MITRE CWE corpus.

CVSS Vector

CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:H/A:H

References

Timeline

Published
July 17, 2026
Last Modified
November 12, 2026
First Seen
September 30, 2026

Related Vulnerabilities