CVE-2022-29201: TensorFlow: QuantizedConv2D null deref crashes model server

MEDIUMPoC AVAILABLECISA: TRACK*

Published May 20, 2022

CISO Take

A null pointer dereference in TensorFlow's QuantizedConv2D op allows any local user with low privileges to crash the TF runtime by passing empty input tensors, causing a denial of service. Patch immediately to TF 2.9.0, 2.8.1, 2.7.2, or 2.6.4 — this is a one-line exploit against unpatched inference servers. Risk is contained to availability; no data exfiltration or code execution confirmed.

What is the risk?

MEDIUM risk. CVSS 5.5 with local attack vector limits blast radius, but in multi-tenant ML inference environments (Jupyter hubs, shared training clusters, model serving APIs that expose raw TF ops) the effective attack surface widens significantly. No active exploitation in the wild and not in CISA KEV. The low attack complexity means any authenticated cluster user can weaponize this trivially once the CVE is public. Quantized models are common in edge/embedded and mobile deployments, broadening the exposure surface beyond data centers.

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
TensorFlow	pip	—	No patch
201.3KOpenSSF 7.63.4K dependentsPushed 5d ago4% patched~1372d to patchFull package profile →

Do you use TensorFlow? You're affected.

How severe is it?

CVSS 3.1

5.5 / 10

EPSS

0.3%

chance of exploitation in 30 days

Higher than 28% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

Exploit Available

Exploitation: MEDIUM

Sophistication

Trivial

Exploitation Confidence

medium

○CISA SSVC: Public PoC

○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?

AVLocal

ACLow

PRLow

UINone

SUnchanged

CNone

INone

AHigh

What should I do?

5 steps

Patch: Upgrade TensorFlow to >=2.9.0, >=2.8.1, >=2.7.2, or >=2.6.4 immediately.
Workaround if patching is blocked: Add input validation middleware to reject requests with empty or zero-dimension tensors before they reach TF ops.
Detection: Monitor for unexpected TF process crashes or SIGSEGV/SIGABRT signals in model serving logs; alert on repeated process restarts in serving infrastructure.
Restrict access: Enforce least privilege on who can submit inference requests to endpoints exposing raw TF ops — prefer typed APIs over passthrough op execution.
Inventory: Audit whether any model in production uses QuantizedConv2D ops and prioritize those services for patching.

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?

DoSFrameworkInferenceAML.T0010.001- AI Software AML.T0029- Denial of AI Service AML.T0043- Craft Adversarial Data AML.T0049- Exploit Public-Facing Application

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act

Article 15-Accuracy, robustness and cybersecurity

ISO 42001

A.9.3-Technical testing of AI systems

NIST AI RMF

GOVERN 6.2-Policies and procedures for AI risk management are in placeMANAGE 2.4-Residual risks from AI system operation are monitored and managed

Frequently Asked Questions

What is CVE-2022-29201?

Is CVE-2022-29201 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2022-29201, increasing the risk of exploitation.

How to fix CVE-2022-29201?

1. Patch: Upgrade TensorFlow to >=2.9.0, >=2.8.1, >=2.7.2, or >=2.6.4 immediately. 2. Workaround if patching is blocked: Add input validation middleware to reject requests with empty or zero-dimension tensors before they reach TF ops. 3. Detection: Monitor for unexpected TF process crashes or SIGSEGV/SIGABRT signals in model serving logs; alert on repeated process restarts in serving infrastructure. 4. Restrict access: Enforce least privilege on who can submit inference requests to endpoints exposing raw TF ops — prefer typed APIs over passthrough op execution. 5. Inventory: Audit whether any model in production uses QuantizedConv2D ops and prioritize those services for patching.

What systems are affected by CVE-2022-29201?

This vulnerability affects the following AI/ML architecture patterns: model serving, training pipelines, edge/embedded inference, quantization-aware training pipelines, multi-tenant ML platforms.

What is the CVSS score for CVE-2022-29201?

CVE-2022-29201 has a CVSS v3.1 base score of 5.5 (MEDIUM). The EPSS exploitation probability is 0.34%.

What is the AI security impact?

Affected AI Architectures

model servingtraining pipelinesedge/embedded inferencequantization-aware training pipelinesmulti-tenant ML platforms

MITRE ATLAS Techniques

AML.T0010.001AI Software

AML.T0029Denial of AI Service

AML.T0043Craft Adversarial Data

AML.T0049Exploit Public-Facing Application

Compliance Controls Affected

EU AI Act: Article 15

ISO 42001: A.9.3

NIST AI RMF: GOVERN 6.2, MANAGE 2.4

What are the technical details?

Original Advisory

TensorFlow is an open source platform for machine learning. Prior to versions 2.9.0, 2.8.1, 2.7.2, and 2.6.4, the implementation of `tf.raw_ops.QuantizedConv2D` does not fully validate the input arguments. In this case, references get bound to `nullptr` for each argument that is empty. Versions 2.9.0, 2.8.1, 2.7.2, and 2.6.4 contain a patch for this issue.

Exploitation Scenario

An adversary with a valid account on a shared ML platform (e.g., internal Jupyter hub or multi-tenant KServe deployment) identifies that a production model uses quantized convolution layers. They craft an inference request passing an empty tensor as the filter or input argument to tf.raw_ops.QuantizedConv2D. TensorFlow fails to validate the argument, binds a nullptr, and dereferences it during kernel execution, causing an immediate process crash. In a model serving context without auto-restart, the endpoint becomes unavailable. With auto-restart and no rate limiting, the attacker replays the request continuously to maintain a denial-of-service condition against all models sharing that serving instance.

Weaknesses (CWE)

CWE-20Improper Input Validation CWE-476NULL Pointer Dereference

CWE-20 — Improper Input Validation: The product receives input or data, but it does not validate or incorrectly validates that the input has the properties that are required to process the data safely and correctly.

[Architecture and Design] Consider using language-theoretic security (LangSec) techniques that characterize inputs using a formal language and build "recognizers" for that language. This effectively requires parsing to be a distinct layer that effectively enforces a boundary between raw input and internal data representations, instead of allowing parser code to be scattered throughout the program, where it could be subject to errors or inconsistencies that create weaknesses. [REF-1109] [REF-1110] [REF-1111]
[Architecture and Design] Use an input validation framework such as Struts or the OWASP ESAPI Validation API. Note that using a framework does not automatically address all input validation problems; be mindful of weaknesses that could arise from misusing the framework itself (CWE-1173).

Source: MITRE CWE corpus.