CVE-2022-35990: TensorFlow DoS — HIGH

CISO Take

A remotely exploitable, unauthenticated DoS exists in TensorFlow's quantization-aware training gradient op — a malformed input tensor rank triggers a fatal assertion, crashing the process. If you expose TensorFlow serving or training endpoints to untrusted or user-supplied inputs, patch immediately to TF 2.10.0, 2.9.1, 2.8.1, or 2.7.2. No workaround exists; the fix is patching.

What is the risk?

CVSS 7.5 High with network-accessible, zero-auth, low-complexity exploitation raises the operational bar significantly for AI/ML teams running TF-based inference APIs. The DoS-only impact (no data exfiltration or code execution) tempers urgency, and the specific op affected (fake_quant_with_min_max_vars_per_channel_gradient) is more common in QAT training pipelines than pure inference deployments. Real-world risk is highest for orgs exposing MLOps platforms, notebook servers, or training APIs to multi-tenant or internet-facing environments. Not in CISA KEV; no known active exploitation.

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
TensorFlow	pip	—	No patch
195.8K OpenSSF 7.1 3.7K dependents Pushed 3d ago 4% patched ~1372d to patch Full package profile →

Do you use TensorFlow? You're affected.

How severe is it?

CVSS 3.1

7.5 / 10

EPSS

0.4%

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

○ 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?

AV Network

AC Low

PR None

UI None

S Unchanged

C None

I None

A High

What should I do?

5 steps

Patch: Upgrade to TensorFlow 2.10.0, or apply cherrypicks to 2.9.1, 2.8.1, or 2.7.2. No other workaround exists per upstream.
Defense-in-depth: Validate tensor rank (must be 1) for min/max inputs before reaching fake_quant ops — add input schema validation at API layer.
Network controls: Restrict access to TF Serving and training endpoints to authorized clients only; do not expose to public internet without authentication.
Detection: Monitor TF process logs for CHECK failure stack traces referencing fake_quant_with_min_max_vars_per_channel_gradient; alert on unexpected process restarts in serving containers.
Inventory: Identify all TF versions in your ML stack — include transitive dependencies in MLOps tooling (notebooks, AutoML platforms).

What does CISA's SSVC say?

Decision Track

Exploitation none

Automatable No

Technical Impact partial

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

How is it classified?

DoS Framework Inference AML.T0010.001 - AI Software AML.T0029 - Denial of AI Service 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

8.4 - AI system operation and monitoring

NIST AI RMF

MANAGE 2.2 - Mechanisms to sustain AI system operation

OWASP LLM Top 10

LLM04 - Model Denial of Service

Frequently Asked Questions

What is CVE-2022-35990?

A remotely exploitable, unauthenticated DoS exists in TensorFlow's quantization-aware training gradient op — a malformed input tensor rank triggers a fatal assertion, crashing the process. If you expose TensorFlow serving or training endpoints to untrusted or user-supplied inputs, patch immediately to TF 2.10.0, 2.9.1, 2.8.1, or 2.7.2. No workaround exists; the fix is patching.

Is CVE-2022-35990 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2022-35990, increasing the risk of exploitation.

How to fix CVE-2022-35990?

1. Patch: Upgrade to TensorFlow 2.10.0, or apply cherrypicks to 2.9.1, 2.8.1, or 2.7.2. No other workaround exists per upstream. 2. Defense-in-depth: Validate tensor rank (must be 1) for min/max inputs before reaching fake_quant ops — add input schema validation at API layer. 3. Network controls: Restrict access to TF Serving and training endpoints to authorized clients only; do not expose to public internet without authentication. 4. Detection: Monitor TF process logs for CHECK failure stack traces referencing fake_quant_with_min_max_vars_per_channel_gradient; alert on unexpected process restarts in serving containers. 5. Inventory: Identify all TF versions in your ML stack — include transitive dependencies in MLOps tooling (notebooks, AutoML platforms).

What systems are affected by CVE-2022-35990?

This vulnerability affects the following AI/ML architecture patterns: training pipelines, model serving, MLOps platforms, edge/mobile model optimization.

What is the CVSS score for CVE-2022-35990?

CVE-2022-35990 has a CVSS v3.1 base score of 7.5 (HIGH). The EPSS exploitation probability is 0.37%.

What is the AI security impact?

Affected AI Architectures

training pipelinesmodel servingMLOps platformsedge/mobile model optimization

MITRE ATLAS Techniques

AML.T0010.001 AI Software

AML.T0029 Denial of AI Service

AML.T0049 Exploit Public-Facing Application

Compliance Controls Affected

EU AI Act: Article 15

ISO 42001: 8.4

NIST AI RMF: MANAGE 2.2

OWASP LLM Top 10: LLM04

What are the technical details?

Original Advisory

TensorFlow is an open source platform for machine learning. When `tf.quantization.fake_quant_with_min_max_vars_per_channel_gradient` receives input `min` or `max` of rank other than 1, it gives a `CHECK` fail that can trigger a denial of service attack. We have patched the issue in GitHub commit f3cf67ac5705f4f04721d15e485e192bb319feed. 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 running a quantized TensorFlow model behind an inference API sends a crafted POST request to the /v1/models/:predict endpoint, supplying a min or max parameter as a rank-2 (or higher) tensor instead of the expected rank-1 vector. TensorFlow's internal CHECK assertion fires immediately, raising a SIGABRT that kills the serving worker. In a Kubernetes-based deployment this triggers a CrashLoopBackOff, effectively taking the inference endpoint offline. The attack requires no authentication, no ML expertise beyond reading TF API documentation, and can be automated to maintain a persistent DoS condition against the targeted AI service.

Weaknesses (CWE)

CWE-617 Reachable 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] Make sensitive open/close operation non reachable by directly user-controlled data (e.g. open/close resources)
[Implementation] Perform input validation on user data.

Source: MITRE CWE corpus.