CVE-2022-35959 — HIGH (CVSS 7.5) AI Security Vulnerability

CISO Take

A network-accessible denial-of-service flaw in TensorFlow's 3D pooling gradient operator lets unauthenticated attackers crash ML serving endpoints with a single crafted request. Any production TF deployment accepting user-controlled tensor shapes is at risk. No workarounds exist — patch immediately to TF 2.10.0, 2.9.1, 2.8.1, or 2.7.2.

Risk Assessment

High severity (CVSS 7.5) with a trivially exploitable, zero-auth, network-accessible attack path. The overflow-to-CHECK-failure chain requires no ML expertise — just a malformed API call. Impact is strictly availability (DoS); no data exfiltration or code execution surface exists. Risk is concentrated in TensorFlow Serving deployments exposed to untrusted networks or multi-tenant inference APIs. Not in CISA KEV and no active exploitation reported, but the AC:L/PR:N/UI:N vector justifies immediate patching over deferred scheduling.

Affected Systems

Package	Ecosystem	Vulnerable Range	Patched
tensorflow	pip	—	No patch
195.0K OpenSSF 7.2 3.7K dependents Pushed today 4% patched ~1372d to patch Full package profile →

Do you use tensorflow? You're affected.

Severity & Risk

CVSS 3.1

7.5 / 10

EPSS

0.1%

chance of exploitation in 30 days

Higher than 20% 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, CISA SSVC, EPSS, trickest/cve, and Nuclei templates.

Attack Surface

AV Network

AC Low

PR None

UI None

S Unchanged

C None

I None

A High

Recommended Action

5 steps

Patch: Upgrade to TensorFlow 2.10.0, 2.9.1, 2.8.1, or 2.7.2 (fix in commit 9178ac9d).
Input validation: Enforce strict tensor shape schema validation at the API gateway before requests reach the model — reject shapes outside expected bounds, especially for orig_input_shape parameters.
Rate limiting: Apply per-client rate limits on inference endpoints to reduce sustained DoS impact.
Detection: Alert on unexpected TF Serving process crashes or CHECK failure log entries (grep for 'Check failed' in TF logs).
Network segmentation: Do not expose TF Serving directly to the public internet — require authenticated, authorized access to all inference endpoints.

CISA SSVC Assessment

Decision Track

Exploitation none

Automatable No

Technical Impact partial

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

Classification

DoS Framework Inference AML.T0010.001 - AI Software AML.T0029 - Denial of AI Service AML.T0049 - Exploit Public-Facing Application

Compliance Impact

This CVE is relevant to:

EU AI Act

Article 15 - Accuracy, robustness, and cybersecurity

ISO 42001

Clause 8.4 - AI system operation and monitoring

NIST AI RMF

MANAGE 2.2 - Mechanisms to sustain the value of AI systems are maintained

OWASP LLM Top 10

LLM04 - Model Denial of Service

Frequently Asked Questions

What is CVE-2022-35959?

A network-accessible denial-of-service flaw in TensorFlow's 3D pooling gradient operator lets unauthenticated attackers crash ML serving endpoints with a single crafted request. Any production TF deployment accepting user-controlled tensor shapes is at risk. No workarounds exist — patch immediately to TF 2.10.0, 2.9.1, 2.8.1, or 2.7.2.

Is CVE-2022-35959 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2022-35959, increasing the risk of exploitation.

How to fix CVE-2022-35959?

1. Patch: Upgrade to TensorFlow 2.10.0, 2.9.1, 2.8.1, or 2.7.2 (fix in commit 9178ac9d). 2. Input validation: Enforce strict tensor shape schema validation at the API gateway before requests reach the model — reject shapes outside expected bounds, especially for orig_input_shape parameters. 3. Rate limiting: Apply per-client rate limits on inference endpoints to reduce sustained DoS impact. 4. Detection: Alert on unexpected TF Serving process crashes or CHECK failure log entries (grep for 'Check failed' in TF logs). 5. Network segmentation: Do not expose TF Serving directly to the public internet — require authenticated, authorized access to all inference endpoints.

What systems are affected by CVE-2022-35959?

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

What is the CVSS score for CVE-2022-35959?

CVE-2022-35959 has a CVSS v3.1 base score of 7.5 (HIGH). The EPSS exploitation probability is 0.06%.

Technical Details

NVD Description

TensorFlow is an open source platform for machine learning. The implementation of `AvgPool3DGradOp` does not fully validate the input `orig_input_shape`. This results in an overflow that results in a `CHECK` failure which can be used to trigger a denial of service attack. We have patched the issue in GitHub commit 9178ac9d6389bdc54638ab913ea0e419234d14eb. 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 identifies a publicly accessible TF Serving endpoint running a video classification or 3D medical imaging model containing AvgPool3D layers. They submit a gRPC or HTTP inference request with an intentionally malformed orig_input_shape — e.g., an extremely large, zero, or negative tensor dimension — crafted to trigger an integer overflow in AvgPool3DGradOp. The overflow causes a CHECK assertion failure, immediately crashing the TF Serving process. With no authentication required and reproducible results, the adversary can script repeated requests to maintain persistent service disruption, effectively taking down model inference for all downstream applications and users sharing the endpoint.