CVE-2022-35972 — HIGH (CVSS 7.5) AI Security Vulnerability

Q: Is CVE-2022-35972 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2022-35972, increasing the risk of exploitation.

Q: How to fix CVE-2022-35972?

1. PATCH: Upgrade to TensorFlow 2.10.0, 2.9.1, 2.8.1, or 2.7.2 — these contain commit 785d67a. No workarounds exist per the advisory. 2. VALIDATE INPUTS: Add tensor rank validation at API boundaries before passing to TF ops. Reject any request where min_input, max_input, min_bias, or max_bias tensors have rank > 0. 3. PROCESS ISOLATION: Run TF serving processes under supervisord/systemd with auto-restart to minimize downtime from crashes. 4. AUDIT: Inventory all TensorFlow versions in use across inference, training, and evaluation environments — check Dockerfiles, requirements.txt, conda envs. 5. DETECT: Monitor serving processes for unexpected crashes or SIGSEGV signals; anomalous crash rates may indicate active exploitation.

CISO Take

Any TensorFlow deployment serving quantized models over a network endpoint is vulnerable to process crashes via crafted tensor inputs — no credentials required. If you're running TensorFlow < 2.10.0 in model serving, inference APIs, or TFX pipelines, patch immediately. This is a 2022 vulnerability with public patches; unpatched instances represent a maintenance gap, not a zero-day.

Risk Assessment

High exploitability: network-accessible, zero auth, low complexity, no user interaction required. Impact is limited to availability (DoS), with no confidentiality or integrity risk. The attack surface is real in production ML inference APIs that accept external inputs processed by quantized TensorFlow ops. Not in CISA KEV and not known to be actively exploited in the wild, but the trivial exploitation path (just send malformed tensor ranks) lowers the bar significantly. Risk is elevated for organizations running legacy TensorFlow versions in production serving infrastructure.

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 — these contain commit 785d67a. No workarounds exist per the advisory.
VALIDATE INPUTS

Add tensor rank validation at API boundaries before passing to TF ops. Reject any request where min_input, max_input, min_bias, or max_bias tensors have rank > 0.
PROCESS ISOLATION

Run TF serving processes under supervisord/systemd with auto-restart to minimize downtime from crashes.
AUDIT

Inventory all TensorFlow versions in use across inference, training, and evaluation environments — check Dockerfiles, requirements.txt, conda envs.
DETECT

Monitor serving processes for unexpected crashes or SIGSEGV signals; anomalous crash rates may indicate active exploitation.

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

Compliance Impact

This CVE is relevant to:

EU AI Act

Art. 15 - Accuracy, robustness and cybersecurity

ISO 42001

8.4 - AI system operation

NIST AI RMF

GOVERN-5.2 - Organizational risk policies for AI deployment MANAGE-2.4 - Residual risks from third-party AI components

Frequently Asked Questions

What is CVE-2022-35972?

Any TensorFlow deployment serving quantized models over a network endpoint is vulnerable to process crashes via crafted tensor inputs — no credentials required. If you're running TensorFlow < 2.10.0 in model serving, inference APIs, or TFX pipelines, patch immediately. This is a 2022 vulnerability with public patches; unpatched instances represent a maintenance gap, not a zero-day.

Is CVE-2022-35972 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2022-35972, increasing the risk of exploitation.

How to fix CVE-2022-35972?

1. PATCH: Upgrade to TensorFlow 2.10.0, 2.9.1, 2.8.1, or 2.7.2 — these contain commit 785d67a. No workarounds exist per the advisory. 2. VALIDATE INPUTS: Add tensor rank validation at API boundaries before passing to TF ops. Reject any request where min_input, max_input, min_bias, or max_bias tensors have rank > 0. 3. PROCESS ISOLATION: Run TF serving processes under supervisord/systemd with auto-restart to minimize downtime from crashes. 4. AUDIT: Inventory all TensorFlow versions in use across inference, training, and evaluation environments — check Dockerfiles, requirements.txt, conda envs. 5. DETECT: Monitor serving processes for unexpected crashes or SIGSEGV signals; anomalous crash rates may indicate active exploitation.

What systems are affected by CVE-2022-35972?

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

What is the CVSS score for CVE-2022-35972?

CVE-2022-35972 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. If `QuantizedBiasAdd` is given `min_input`, `max_input`, `min_bias`, `max_bias` tensors of a nonzero rank, it results in a segfault that can be used to trigger a denial of service attack. We have patched the issue in GitHub commit 785d67a78a1d533759fcd2f5e8d6ef778de849e0. 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 organization's ML inference API identifies a quantized TensorFlow model endpoint (e.g., a product recommendation or fraud detection model). They craft an HTTP request with tensors of nonzero rank for the min_input/max_input/min_bias/max_bias parameters of a QuantizedBiasAdd operation. The malformed input triggers a segfault in the TF process, crashing the inference server. With no rate limiting or input shape validation, the attacker can automate this to maintain a sustained DoS against the model serving layer — disrupting dependent business logic without needing any credentials or prior system knowledge.