CVE-2022-23580: TensorFlow uncontrolled allocation

CISO Take

Any TensorFlow deployment exposing model inference to authenticated users is vulnerable to targeted availability attacks via crafted tensor inputs. The CVSS 6.5 rating understates operational risk in production ML serving: a low-privilege API consumer can crash your inference service with a single request. Patch immediately to TF 2.8.0, 2.7.1, 2.6.3, or 2.5.3 and enforce input shape validation at the API gateway layer.

What is the risk?

Medium severity by CVSS but operationally high-risk for ML serving infrastructure. Exploitation requires only low privileges (API key or authenticated session), no user interaction, and low complexity — making it accessible to any adversary with inference API access. Impact is confined to availability (no data exfiltration path), but in production AI systems, inference downtime directly translates to service outages and SLA violations. Organizations running multi-tenant ML platforms or exposing TensorFlow serving APIs externally face the highest exposure.

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
TensorFlow	pip	—	No patch
195.8K OpenSSF 7.1 3.7K dependents Pushed 2d ago 4% patched ~1372d to patch Full package profile →

Do you use TensorFlow? You're affected.

How severe is it?

CVSS 3.1

6.5 / 10

EPSS

0.8%

chance of exploitation in 30 days

Higher than 52% 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?

AV Network

AC Low

PR Low

UI None

S Unchanged

C None

I None

A High

What should I do?

6 steps

PATCH

Upgrade to TensorFlow 2.8.0, 2.7.1, 2.6.3, or 2.5.3 — apply the fix from commit 1361fb7e.
VALIDATE INPUT

Enforce strict tensor shape and size limits at the API gateway before requests reach TF runtime; reject any tensor dimension exceeding expected bounds.
RATE LIMIT

Apply per-user/per-key rate limiting on inference endpoints to contain blast radius from abuse.
RESOURCE LIMITS

Configure OOM kill policies and container memory limits on inference pods to enable fast recovery.
DETECT

Monitor for sudden memory spikes or service restarts on inference nodes as indicators of exploitation attempts.
ISOLATE

Run inference services in isolated containers/processes so a crash does not cascade to adjacent services.

What does CISA's SSVC say?

Decision Track*

Exploitation poc

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.T0029 - Denial of AI Service AML.T0034 - Cost Harvesting 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 - AI system operation and monitoring

NIST AI RMF

MANAGE-2.2 - Mechanisms to sustain deployed AI and apply risk mitigations

OWASP LLM Top 10

LLM10:2025 - Unbounded Consumption

Frequently Asked Questions

What is CVE-2022-23580?

Any TensorFlow deployment exposing model inference to authenticated users is vulnerable to targeted availability attacks via crafted tensor inputs. The CVSS 6.5 rating understates operational risk in production ML serving: a low-privilege API consumer can crash your inference service with a single request. Patch immediately to TF 2.8.0, 2.7.1, 2.6.3, or 2.5.3 and enforce input shape validation at the API gateway layer.

Is CVE-2022-23580 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2022-23580, increasing the risk of exploitation.

How to fix CVE-2022-23580?

1. PATCH: Upgrade to TensorFlow 2.8.0, 2.7.1, 2.6.3, or 2.5.3 — apply the fix from commit 1361fb7e. 2. VALIDATE INPUT: Enforce strict tensor shape and size limits at the API gateway before requests reach TF runtime; reject any tensor dimension exceeding expected bounds. 3. RATE LIMIT: Apply per-user/per-key rate limiting on inference endpoints to contain blast radius from abuse. 4. RESOURCE LIMITS: Configure OOM kill policies and container memory limits on inference pods to enable fast recovery. 5. DETECT: Monitor for sudden memory spikes or service restarts on inference nodes as indicators of exploitation attempts. 6. ISOLATE: Run inference services in isolated containers/processes so a crash does not cascade to adjacent services.

What systems are affected by CVE-2022-23580?

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

What is the CVSS score for CVE-2022-23580?

CVE-2022-23580 has a CVSS v3.1 base score of 6.5 (MEDIUM). The EPSS exploitation probability is 0.81%.

What is the AI security impact?

Affected AI Architectures

model servingtraining pipelinesinference endpointsML platform APIs

MITRE ATLAS Techniques

AML.T0029 Denial of AI Service

AML.T0034 Cost Harvesting

AML.T0049 Exploit Public-Facing Application

Compliance Controls Affected

EU AI Act: Article 15

ISO 42001: A.9.3

NIST AI RMF: MANAGE-2.2

OWASP LLM Top 10: LLM10:2025

What are the technical details?

Original Advisory

Tensorflow is an Open Source Machine Learning Framework. During shape inference, TensorFlow can allocate a large vector based on a value from a tensor controlled by the user. The fix will be included in TensorFlow 2.8.0. We will also cherrypick this commit on TensorFlow 2.7.1, TensorFlow 2.6.3, and TensorFlow 2.5.3, as these are also affected and still in supported range.

Exploitation Scenario

An attacker with a valid API key to a TensorFlow Serving endpoint constructs a crafted inference request containing a tensor with an extremely large dimension value in a field that feeds into shape inference. When TF processes the request, shape_inference.cc:788-790 allocates a vector sized by the attacker-controlled value, exhausting available memory. The inference server process crashes or becomes unresponsive. The attacker can repeat this on recovery to maintain a persistent DoS condition against a competitor's AI API, an internal ML platform, or a security-critical AI decision system (e.g., fraud detection, anomaly detection). No exploit code required — a single malformed gRPC or REST inference request suffices.

Weaknesses (CWE)

CWE-1284 Improper Validation of Specified Quantity in Input Primary CWE-400 Uncontrolled Resource Consumption

CWE-1284 — Improper Validation of Specified Quantity in Input: The product receives input that is expected to specify a quantity (such as size or length), but it does not validate or incorrectly validates that the quantity has the required properties.

[Implementation] Assume all input is malicious. Use an "accept known good" input validation strategy, i.e., use a list of acceptable inputs that strictly conform to specifications. Reject any input that does not strictly conform to specifications, or transform it into something that does. When performing input validation, consider all potentially relevant properties, including length, type of input, the full range of acceptable values, missing or extra inputs, syntax, consistency across related fields, and conformance to business rules. As an example of business rule logic, "boat" may be syntactically valid because it only contains alphanumeric characters, but it is not valid if the input is only expected to contain colors such as "red" or "blue." Do not rely exclusively on looking for malicious or malformed inputs. This is likely to miss at least one undesirable input, especially if the code's environment changes. This can give attackers enough room to bypass the intended validation. However, denylis

Source: MITRE CWE corpus.