CVE-2022-36018: TensorFlow: RaggedTensor CHECK fail remote DoS

HIGHPoC AVAILABLE

Published September 16, 2022

CISO Take

Any TensorFlow deployment accepting untrusted RaggedTensor inputs over the network is vulnerable to process crash with zero authentication required. This is a straightforward availability risk for ML inference APIs and serving infrastructure. Patch immediately to TF 2.10.0, 2.9.1, 2.8.1, or 2.7.2 — no workaround exists.

What is the risk?

High severity (CVSS 7.5) with worst-case exploitability characteristics: network-accessible, no privileges, no user interaction, low complexity. The CHECK fail causes immediate process termination, making it a reliable DoS primitive. Risk is highest for ML inference services and model-serving APIs that accept user-controlled tensor data. Reduced risk in air-gapped training environments where inputs are fully controlled. Not in CISA KEV and no evidence of active exploitation, but the trivial exploit path warrants prompt patching.

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
TensorFlow	pip	—	No patch
201.3KOpenSSF 7.63.4K dependentsPushed 6d ago4% patched~1372d to patchFull package profile →

Do you use TensorFlow? You're affected.

How severe is it?

CVSS 3.1

7.5 / 10

EPSS

0.5%

chance of exploitation in 30 days

Higher than 40% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

Exploit Available

Exploitation: MEDIUM

Sophistication

Trivial

Exploitation Confidence

medium

○Public PoC indexed (trickest/cve)

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?

AVNetwork

ACLow

PRNone

UINone

SUnchanged

CNone

INone

AHigh

What should I do?

5 steps

PATCH
Upgrade to TensorFlow 2.10.0, or cherry-pick backports: 2.9.1, 2.8.1, 2.7.2 (commit 88f93dfe).
VALIDATE INPUTS
At inference endpoints, validate tensor ranks before processing — reject rt_nested_splits inputs with rank != 1.
ISOLATE
Run TF Serving in isolated containers/VMs; a crash should not cascade to other services.
MONITOR
Alert on abnormal TF Serving process restarts or crash loops — these may indicate active exploitation attempts.
AUDIT
Identify all internal and external-facing TensorFlow endpoints accepting RaggedTensor inputs.

What does CISA's SSVC say?

DecisionTrack

Exploitationnone

AutomatableNo

Technical Impactpartial

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

How is it classified?

DoSFrameworkInferenceAML.T0029- Denial of AI Service AML.T0034- Cost Harvesting AML.T0049- Exploit Public-Facing Application

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act

Article 15-Accuracy, robustness and cybersecurityArticle 9-Risk management system

ISO 42001

A.6.2.6-AI system operational monitoring

NIST AI RMF

MANAGE-2.2-Mechanisms to sustain the value of deployed AI are evaluated and applied

Frequently Asked Questions

What is CVE-2022-36018?

Is CVE-2022-36018 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2022-36018, increasing the risk of exploitation.

How to fix CVE-2022-36018?

1. PATCH: Upgrade to TensorFlow 2.10.0, or cherry-pick backports: 2.9.1, 2.8.1, 2.7.2 (commit 88f93dfe). 2. VALIDATE INPUTS: At inference endpoints, validate tensor ranks before processing — reject rt_nested_splits inputs with rank != 1. 3. ISOLATE: Run TF Serving in isolated containers/VMs; a crash should not cascade to other services. 4. MONITOR: Alert on abnormal TF Serving process restarts or crash loops — these may indicate active exploitation attempts. 5. AUDIT: Identify all internal and external-facing TensorFlow endpoints accepting RaggedTensor inputs.

What systems are affected by CVE-2022-36018?

This vulnerability affects the following AI/ML architecture patterns: model serving, inference APIs, training pipelines, NLP processing pipelines.

What is the CVSS score for CVE-2022-36018?

CVE-2022-36018 has a CVSS v3.1 base score of 7.5 (HIGH). The EPSS exploitation probability is 0.46%.

What is the AI security impact?

Affected AI Architectures

model servinginference APIstraining pipelinesNLP processing pipelines

MITRE ATLAS Techniques

AML.T0029Denial of AI Service

AML.T0034Cost Harvesting

AML.T0049Exploit Public-Facing Application

Compliance Controls Affected

EU AI Act: Article 15, Article 9

ISO 42001: A.6.2.6

NIST AI RMF: MANAGE-2.2

What are the technical details?

Original Advisory

TensorFlow is an open source platform for machine learning. If `RaggedTensorToVariant` is given a `rt_nested_splits` list that contains tensors of ranks other than one, it results in a `CHECK` fail that can be used to trigger a denial of service attack. We have patched the issue in GitHub commit 88f93dfe691563baa4ae1e80ccde2d5c7a143821. The fix will be included in TensorFlow 2.10.0. We will also cherrypick this commit on TensorFlow 2.9.1, TensorFlow 2.8.1, and TensorFlow 2.7.2, as these are also affected and still in supported range. There are no known workarounds for this issue.

Exploitation Scenario

An adversary targeting an NLP inference API (e.g., a text classification or sequence labeling service built on TF Serving) sends a crafted gRPC request with a RaggedTensor payload where rt_nested_splits contains a 2D tensor instead of the expected 1D tensor. TensorFlow's RaggedTensorToVariant hits the CHECK assertion, immediately killing the serving process. On most deployment configurations this restarts the container, but repeated requests at low volume can maintain a persistent availability outage with minimal infrastructure cost to the attacker. No ML expertise required — only knowledge of the TF Serving gRPC API schema.

Weaknesses (CWE)

CWE-617Reachable AssertionPrimary CWE-617Reachable Assertion

CWE-617 — Reachable Assertion: The product contains an assert() or similar statement that can be triggered by an attacker, which leads to an application exit or other behavior that is more severe than necessary.

[Implementation] If the assertion statement is required (e.g., for security or reliability), change the invoking code to avoid the conditions that trigger the assertion.
[Implementation] Consider changing the assertion to some other less "fatal" error. This might not be easy to implement, since other code may be needed to handle the error.

Source: MITRE CWE corpus.