CVE-2026-42009: vLLM DTLS packet-reorder bug DoS hits

CISO Take

A flaw in gnutls' DTLS packet-reordering comparator fails to handle duplicate sequence numbers, leading to unstable ordering or undefined behavior that a remote, unauthenticated attacker can trigger over the network to cause a denial of service (CVSS 7.5, AV:N/AC:L/PR:N/UI:N). This isn't an AI-specific attack technique, but it lands squarely on AI infrastructure: Red Hat's AI Inference Server images (rhaiis/vllm-rocm-rhel9, rhaiis/model-opt-cuda-rhel9) bundle the vulnerable library, and gnutls itself has 129 downstream dependents and a package risk score of 61/100 with 82 other CVEs on record. The EPSS score of 1.3% (top 32nd percentile) is modest, there's no CISA KEV listing, no public exploit, and no Nuclei template, so urgency is moderate rather than critical. Patch via the relevant RHSA errata (RHSA-2026:13274 and related builds) to update gnutls/libtasn1 across affected images, prioritizing any inference-serving hosts that terminate DTLS, and monitor for unexpected crash-restart patterns on DTLS-facing services as a detection signal in the interim.

Sources: NVD CISA KEV EPSS ATLAS access.redhat.com

What is the risk?

CVSS 7.5 (high) reflects a network-exploitable, unauthenticated, low-complexity DoS with no user interaction required — a favorable exploit profile on paper. However, real-world exploitation likelihood is tempered: EPSS sits at just 1.3% (top 32% percentile, not top-tier), there is no CISA KEV entry, no public exploit code, and no Nuclei scanner template, indicating this has not attracted active attacker interest. The residual risk driver is exposure and blast radius rather than active exploitation: gnutls is a foundational TLS/DTLS library with 129 downstream dependents and a history of 82 other CVEs (package risk score 61/100), and it ships inside Red Hat's AI Inference Server container images, meaning any DTLS-terminating AI inference deployment inherits this exposure until patched.

How does the attack unfold?

Reconnaissance

Attacker identifies a network-reachable DTLS endpoint on infrastructure using gnutls, potentially including AI inference server images.

AML.T0006

Exploitation

Attacker sends crafted DTLS packets with duplicate sequence numbers to trigger the flawed comparator logic in gnutls' packet-reordering routine.

AML.T0049

Impact

The service crashes or hangs due to undefined behavior, causing a denial of service that disrupts availability of the affected AI inference or supporting system.

AML.T0029

Reconnaissance

Attacker identifies a network-reachable DTLS endpoint on infrastructure using gnutls, potentially including AI inference server images.

AML.T0006

Exploitation

Attacker sends crafted DTLS packets with duplicate sequence numbers to trigger the flawed comparator logic in gnutls' packet-reordering routine.

AML.T0049

Impact

The service crashes or hangs due to undefined behavior, causing a denial of service that disrupts availability of the affected AI inference or supporting system.

AML.T0029

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
vLLM	pip	—	No patch
89.7K 128 dependents Pushed 2d ago 24% patched ~52d to patch Full package profile →
discovery/discovery-server-rhel9	—	—	No patch
discovery/discovery-ui-rhel9	—	—	No patch
gnutls	—	—	No patch
gnutls-main	—	—	No patch
libtasn1	—	—	No patch
rhaiis/model-opt-cuda-rhel9	—	—	No patch
rhcos	—	—	No patch
rhui5/cds-rhel9	—	—	No patch
rhui5/haproxy-rhel9	—	—	No patch
rhui5/installer-rhel9	—	—	No patch
rhui5/rhua-rhel9	—	—	No patch

How severe is it?

CVSS 3.1

7.5 / 10

EPSS

1.3%

chance of exploitation in 30 days

Higher than 69% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

No known exploitation

Sophistication

Trivial

What is the attack surface?

AV Network

AC Low

PR None

UI None

S Unchanged

C None

I None

A High

What should I do?

1 step

Patch gnutls and libtasn1 to the fixed versions referenced in the Red Hat errata (RHSA-2026:13274, 20611-20613, 26319, 26409, 29197, 29794, 30004, 30849) across all affected products, prioritizing rhaiis inference-server images and rhcos hosts. Rebuild and redeploy affected container images (rhaiis/vllm-rocm-rhel9, rhaiis/model-opt-cuda-rhel9) rather than patching in place. If immediate patching isn't feasible, restrict network exposure of DTLS-terminating services via firewall/network policy and disable DTLS where not strictly required. For detection, monitor for abnormal crash-restart loops or service interruptions on DTLS-facing endpoints, and inspect traffic for repeated packets carrying duplicate DTLS sequence numbers as a leading indicator of exploitation attempts.

What does CISA's SSVC say?

Decision Track

Exploitation none

Automatable Yes

Technical Impact partial

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

How is it classified?

DoS Inference AML.T0029 - Denial of AI Service AML.T0049 - Exploit Public-Facing Application

Which compliance frameworks are affected?

This CVE is relevant to:

ISO 42001

Annex A.6.2.6 - Operation and monitoring of the AI system

NIST AI RMF

MEASURE 2.7 - AI system performance is evaluated for deployment conditions, including exposure to security threats and adversarial actions

Frequently Asked Questions

What is CVE-2026-42009?

A flaw in gnutls' DTLS packet-reordering comparator fails to handle duplicate sequence numbers, leading to unstable ordering or undefined behavior that a remote, unauthenticated attacker can trigger over the network to cause a denial of service (CVSS 7.5, AV:N/AC:L/PR:N/UI:N). This isn't an AI-specific attack technique, but it lands squarely on AI infrastructure: Red Hat's AI Inference Server images (rhaiis/vllm-rocm-rhel9, rhaiis/model-opt-cuda-rhel9) bundle the vulnerable library, and gnutls itself has 129 downstream dependents and a package risk score of 61/100 with 82 other CVEs on record. The EPSS score of 1.3% (top 32nd percentile) is modest, there's no CISA KEV listing, no public exploit, and no Nuclei template, so urgency is moderate rather than critical. Patch via the relevant RHSA errata (RHSA-2026:13274 and related builds) to update gnutls/libtasn1 across affected images, prioritizing any inference-serving hosts that terminate DTLS, and monitor for unexpected crash-restart patterns on DTLS-facing services as a detection signal in the interim.

Is CVE-2026-42009 actively exploited?

No confirmed active exploitation of CVE-2026-42009 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-42009?

Patch gnutls and libtasn1 to the fixed versions referenced in the Red Hat errata (RHSA-2026:13274, 20611-20613, 26319, 26409, 29197, 29794, 30004, 30849) across all affected products, prioritizing rhaiis inference-server images and rhcos hosts. Rebuild and redeploy affected container images (rhaiis/vllm-rocm-rhel9, rhaiis/model-opt-cuda-rhel9) rather than patching in place. If immediate patching isn't feasible, restrict network exposure of DTLS-terminating services via firewall/network policy and disable DTLS where not strictly required. For detection, monitor for abnormal crash-restart loops or service interruptions on DTLS-facing endpoints, and inspect traffic for repeated packets carrying duplicate DTLS sequence numbers as a leading indicator of exploitation attempts.

What systems are affected by CVE-2026-42009?

This vulnerability affects the following AI/ML architecture patterns: model serving, network/transport layer (DTLS), AI inference infrastructure.

What is the CVSS score for CVE-2026-42009?

CVE-2026-42009 has a CVSS v3.1 base score of 7.5 (HIGH). The EPSS exploitation probability is 1.33%.

What is the AI security impact?

Affected AI Architectures

model servingnetwork/transport layer (DTLS)AI inference infrastructure

MITRE ATLAS Techniques

AML.T0029 Denial of AI Service

AML.T0049 Exploit Public-Facing Application

Compliance Controls Affected

ISO 42001: Annex A.6.2.6

NIST AI RMF: MEASURE 2.7

What are the technical details?

Original Advisory

A flaw was found in gnutls. A remote attacker could exploit an issue in the Datagram Transport Layer Security (DTLS) packet reordering logic. The comparator function, responsible for ordering DTLS packets by sequence numbers, did not correctly handle packets with duplicate sequence numbers. This could lead to unstable packet ordering or undefined behavior, resulting in a denial of service.

Exploitation Scenario

An attacker with network access to a DTLS-terminating endpoint on an AI inference deployment (for example, a control-plane or telemetry channel on a Red Hat AI Inference Server host) sends a stream of crafted DTLS packets containing duplicate sequence numbers. The vulnerable comparator in the packet-reordering logic mishandles these duplicates, producing unstable ordering or undefined behavior in the gnutls library, which crashes or hangs the process handling the connection — interrupting availability of the inference service until it is restarted or recovered.

Weaknesses (CWE)

CWE-475 Undefined Behavior for Input to API

CWE-475 — Undefined Behavior for Input to API: The behavior of this function is undefined unless its control parameter is set to a specific value.

Source: MITRE CWE corpus.