CVE-2022-41899: TensorFlow: SdcaOptimizer DoS via malformed tensor rank

HIGHPoC AVAILABLECISA: TRACK*

Published November 18, 2022

CISO Take

A remotely exploitable, no-auth-required crash in TensorFlow's SDCA optimizer can take down any ML inference service that accepts external feature inputs. Any production deployment exposing a TensorFlow serving endpoint backed by linear/logistic models trained with SdcaOptimizer is at risk of a full availability loss. Patch immediately to TF 2.11, 2.10.1, 2.9.3, or 2.8.4, and add input rank validation at API boundaries.

What is the risk?

CVSS 7.5 High with AV:N/AC:L/PR:N/UI:N — worst-case exploitability profile for a DoS. No credentials, no user interaction, no complexity barrier: any unauthenticated caller who can reach the inference endpoint can trigger the crash. Impact is pure availability (C:N/I:N/A:H), so no data exposure, but repeated crashes can constitute sustained service disruption. Not in CISA KEV and no confirmed wild exploitation as of enrichment date, which moderates urgency slightly for air-gapped or internal-only deployments.

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 43% 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?

AVNetwork

ACLow

PRNone

UINone

SUnchanged

CNone

INone

AHigh

What should I do?

5 steps

Patch: upgrade to TF 2.11, 2.10.1, 2.9.3, or 2.8.4 (commit 80ff197d).
Validate input tensor rank at API boundary before forwarding to model: reject any request where dense_features or example_state_data rank ≠ 2 with a 400 error.
Run TF Serving behind an API gateway that enforces schema validation on input shape/rank.
Deploy process supervisors (systemd, Kubernetes liveness probes) to auto-restart crashed serving processes — reduces blast radius while patching.
Detection: monitor for abrupt serving process exits and sudden spike in 5xx errors from the inference endpoint as indicators of active exploitation attempts.

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.T0006- Active Scanning 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 (High-Risk AI)

ISO 42001

A.6.2.5-AI System Availability and Resilience

NIST AI RMF

MANAGE-2.4-Risk Monitoring and Incident Response for AI SystemsMAP-5.1-Likelihood and Impact of AI Risks

Frequently Asked Questions

What is CVE-2022-41899?

Is CVE-2022-41899 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2022-41899, increasing the risk of exploitation.

How to fix CVE-2022-41899?

1. Patch: upgrade to TF 2.11, 2.10.1, 2.9.3, or 2.8.4 (commit 80ff197d). 2. Validate input tensor rank at API boundary before forwarding to model: reject any request where dense_features or example_state_data rank ≠ 2 with a 400 error. 3. Run TF Serving behind an API gateway that enforces schema validation on input shape/rank. 4. Deploy process supervisors (systemd, Kubernetes liveness probes) to auto-restart crashed serving processes — reduces blast radius while patching. 5. Detection: monitor for abrupt serving process exits and sudden spike in 5xx errors from the inference endpoint as indicators of active exploitation attempts.

What systems are affected by CVE-2022-41899?

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

What is the CVSS score for CVE-2022-41899?

CVE-2022-41899 has a CVSS v3.1 base score of 7.5 (HIGH). The EPSS exploitation probability is 0.49%.

What is the AI security impact?

Affected AI Architectures

model servingtraining pipelinesML inference APIs

MITRE ATLAS Techniques

AML.T0006Active Scanning

AML.T0029Denial of AI Service

AML.T0049Exploit Public-Facing Application

Compliance Controls Affected

EU AI Act: Article 15

ISO 42001: A.6.2.5

NIST AI RMF: MANAGE-2.4, MAP-5.1

What are the technical details?

Original Advisory

TensorFlow is an open source platform for machine learning. Inputs `dense_features` or `example_state_data` not of rank 2 will trigger a `CHECK` fail in `SdcaOptimizer`. We have patched the issue in GitHub commit 80ff197d03db2a70c6a111f97dcdacad1b0babfa. The fix will be included in TensorFlow 2.11. We will also cherrypick this commit on TensorFlow 2.10.1, 2.9.3, and TensorFlow 2.8.4, as these are also affected and still in supported range.

Exploitation Scenario

Adversary identifies a public or semi-public TensorFlow Serving endpoint via active scanning or documentation disclosure. They determine the model uses SdcaOptimizer (inferrable from model metadata, API schema, or error messages leaking TF op names). They craft a predict request where dense_features has rank 1 (a 1-D tensor instead of 2-D matrix). TensorFlow's CHECK macro fires inside sdca_internal.cc, raising a fatal signal that kills the serving process. With no restart automation, the endpoint goes dark. With restart automation, adversary loops the request to maintain denial of service with trivial request volume — no flood required, one malformed request per restart cycle suffices.

Weaknesses (CWE)

CWE-617Reachable AssertionPrimary CWE-20Improper Input Validation

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.