CVE-2026-58010: GLib: off-by-one read in GVariant causes DoS

MEDIUM
Published June 30, 2026
CISO Take

CVE-2026-58010 is an off-by-one bounds check error in GLib's GVariant parser (gvs_tuple_is_normal) that lets a network-supplied serialized payload trigger a 1-byte out-of-bounds read, leaking one byte of memory or crashing the process when the read crosses a page boundary. GLib is a ubiquitous base library, and this CVE reaches AI infrastructure only indirectly through Red Hat AI container images (vLLM CUDA/ROCm, model-opt, and the rhai base images for CPU, CUDA, ROCm, TPU and Spyre), which bundle the vulnerable glib2. Exploitation likelihood is low: EPSS is 0.85% (top 43%), there is no public exploit, no Nuclei template, it is not in CISA KEV, and CISA SSVC says TRACK. The practical risk is a crash of an inference or serving container if it parses untrusted GVariant data, which is uncommon in typical LLM serving paths. Apply the Red Hat errata (RHSA-2026:49512 and related) through your normal image rebuild cycle and prioritize internet-exposed inference endpoints.

Sources:NVDEPSSCISA SSVCRed Hat Security AdvisoriesATLAS

What is the risk?

Low-to-medium overall. CVSS 6.5 reflects network reachability with no privileges or interaction, but impact is limited to a 1-byte information disclosure and a conditional DoS. Exploitation requires a code path that feeds attacker-controlled serialized GVariant data to GLib, which is rare in AI serving stacks. No known exploitation, no public PoC, and SSVC is TRACK. The AI relevance is as a transitive base-image dependency, not a flaw in an AI framework.

How does the attack unfold?

Initial Access
Attacker sends a crafted serialized GVariant payload over the network to a service in an AI container that parses it with GLib.
AML.T0049
Exploitation
The off-by-one check in gvs_tuple_is_normal causes a 1-byte out-of-bounds read during alignment padding validation.
Impact
The read crosses a page boundary and crashes the process, interrupting inference serving, or leaks one byte of adjacent memory.
AML.T0029

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 →
vLLMpip—No patch
96.9K93 dependentsPushed 3d ago26% patched~37d to patchFull package profile →
GLib——No patch
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
discovery/discovery-server-rhel9——No patch
discovery/discovery-ui-rhel9——No patch
glib2——No patch
mingw-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-kubernetes-tp-rhel9——No patch
rhui5/cds-rhel9——No patch
rhui5/haproxy-rhel9——No patch
rhui5/installer-rhel9——No patch
rhui5/installer-tp-rhel9——No patch
rhui5/rhua-rhel9——No patch
rhui5/rhua-tp-rhel9——No patch

How severe is it?

CVSS 3.1
6.5 / 10
EPSS
0.5%
chance of exploitation in 30 days
Higher than 43% of all CVEs
Exploitation Status
No known exploitation
Sophistication
Moderate

What is the attack surface?

AVACPRUISCIA
AVNetwork
ACLow
PRNone
UINone
SUnchanged
CLow
INone
ALow

What should I do?

1 step
  1. 1) Pull and rebuild images with the fixed glib2 package from the relevant Red Hat errata (RHSA-2026:49512, 55440, 57015, 58981, 61766, 61783, 63135, 63138, 63140, 65762) matching your image and RHEL stream. 2) Update RHAIIS and rhai base images to the latest patched tags and redeploy vLLM and model-optimization workloads. 3) Run an SBOM or image scan (Trivy, Grype, or Red Hat Clair) to find glib2 across all AI images, including custom images built from rhai base images. 4) Restrict network exposure of services that parse GVariant or D-Bus data and avoid exposing such interfaces to untrusted clients. 5) Monitor for repeated unexplained container crashes or segfaults in inference pods, and set restart limits and alerting in Kubernetes.

What does CISA's SSVC say?

DecisionTrack
Exploitationnone
AutomatableYes
Technical Impactpartial

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

How is it classified?

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act
Art. 15-Accuracy, robustness and cybersecurity
ISO 42001
A.6.2.6-AI system operation and monitoring
NIST AI RMF
MANAGE 2.3-Procedures to respond to and recover from unknown risks
OWASP LLM Top 10
LLM03-Supply Chain

Frequently Asked Questions

What is CVE-2026-58010?

CVE-2026-58010 is an off-by-one bounds check error in GLib's GVariant parser (gvs_tuple_is_normal) that lets a network-supplied serialized payload trigger a 1-byte out-of-bounds read, leaking one byte of memory or crashing the process when the read crosses a page boundary. GLib is a ubiquitous base library, and this CVE reaches AI infrastructure only indirectly through Red Hat AI container images (vLLM CUDA/ROCm, model-opt, and the rhai base images for CPU, CUDA, ROCm, TPU and Spyre), which bundle the vulnerable glib2. Exploitation likelihood is low: EPSS is 0.85% (top 43%), there is no public exploit, no Nuclei template, it is not in CISA KEV, and CISA SSVC says TRACK. The practical risk is a crash of an inference or serving container if it parses untrusted GVariant data, which is uncommon in typical LLM serving paths. Apply the Red Hat errata (RHSA-2026:49512 and related) through your normal image rebuild cycle and prioritize internet-exposed inference endpoints.

Is CVE-2026-58010 actively exploited?

No confirmed active exploitation of CVE-2026-58010 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-58010?

1) Pull and rebuild images with the fixed glib2 package from the relevant Red Hat errata (RHSA-2026:49512, 55440, 57015, 58981, 61766, 61783, 63135, 63138, 63140, 65762) matching your image and RHEL stream. 2) Update RHAIIS and rhai base images to the latest patched tags and redeploy vLLM and model-optimization workloads. 3) Run an SBOM or image scan (Trivy, Grype, or Red Hat Clair) to find glib2 across all AI images, including custom images built from rhai base images. 4) Restrict network exposure of services that parse GVariant or D-Bus data and avoid exposing such interfaces to untrusted clients. 5) Monitor for repeated unexplained container crashes or segfaults in inference pods, and set restart limits and alerting in Kubernetes.

What systems are affected by CVE-2026-58010?

This vulnerability affects the following AI/ML architecture patterns: model serving, LLM inference containers, training pipelines, containerized AI base images.

What is the CVSS score for CVE-2026-58010?

CVE-2026-58010 has a CVSS v3.1 base score of 6.5 (MEDIUM). The EPSS exploitation probability is 0.49%.

What is the AI security impact?

Affected AI Architectures

model servingLLM inference containerstraining pipelinescontainerized AI base images

MITRE ATLAS Techniques

AML.T0010.001AI Software
AML.T0029Denial of AI Service

Compliance Controls Affected

EU AI Act: Art. 15
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 GLib. An off-by-one error can occur in the gvs_tuple_is_normal function in the glib/gvariant-serialiser.c file when doing an alignment padding check because the bounds check uses > instead of >=, causing an out-of-bounds read of only 1 byte. This issue can cause a minor information disclosure of 1 byte and a denial of service when the out-of-bounds read crosses a page boundary.

Exploitation Scenario

An attacker identifies an internet-exposed service inside a vLLM or model-optimization container that deserializes GVariant data from the network, for example through a GIO-based helper or a sidecar component. They send a crafted serialized tuple whose alignment padding triggers the off-by-one in gvs_tuple_is_normal. If the extra byte lies on a page boundary, the process faults and the inference worker crashes, interrupting model serving until the pod restarts. Repeating the request can sustain a denial of service against the AI endpoint. The information-disclosure side is a single byte and has little practical value.

Weaknesses (CWE)

CWE-126 — Buffer Over-read: The product reads from a buffer using buffer access mechanisms such as indexes or pointers that reference memory locations after the targeted buffer.

Source: MITRE CWE corpus.

CVSS Vector

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

References

Timeline

Published
June 30, 2026
Last Modified
November 10, 2026
First Seen
September 30, 2026

Related Vulnerabilities