CVE-2022-23577: TensorFlow: null pointer deref crashes model loader
MEDIUMPoC AVAILABLECISA: TRACK*A low-privilege network attacker can crash TensorFlow serving infrastructure by triggering a null pointer dereference in the SavedModel loader's GetInitOp function. This is a denial-of-service risk for any internet-facing or internally-exposed TensorFlow inference endpoint. Patch immediately to TF 2.8.0, 2.7.1, 2.6.3, or 2.5.3; restrict serving endpoint access to trusted principals in the interim.
What is the risk?
Medium risk overall, but operationally significant for organizations running TensorFlow model serving at scale. CVSS 6.5 with network vector and low privileges required makes it trivially exploitable by any authenticated user or compromised service account. Impact is limited to availability (no data exfiltration or code execution possible), but repeated crashes in production inference pipelines could cause meaningful business disruption. Not in CISA KEV and no known active exploitation.
What systems are affected?
| Package | Ecosystem | Vulnerable Range | Patched |
|---|---|---|---|
| TensorFlow | pip | — | No patch |
Do you use TensorFlow? You're affected.
How severe is it?
What is the attack surface?
What should I do?
5 stepsPatch: Upgrade TensorFlow to 2.8.0, or apply cherry-pick backports to 2.7.1, 2.6.3, or 2.5.3.
Network hardening: Place TensorFlow serving endpoints behind authenticated API gateways; do not expose raw TF Serving ports to untrusted networks.
Process isolation: Run TF Serving workers in containers with auto-restart policies so crashes auto-recover (reduces blast radius while patching).
Detection: Alert on abnormal crash/restart rates in model serving containers; log null pointer/segfault signals from TF processes.
If running on unpatched versions, consider disabling dynamic SavedModel loading from untrusted sources.
What does CISA's SSVC say?
Source: CISA Vulnrichment (SSVC v2.0). Decision based on the CISA Coordinator decision tree.
How is it classified?
Which compliance frameworks are affected?
This CVE is relevant to:
Frequently Asked Questions
What is CVE-2022-23577?
A low-privilege network attacker can crash TensorFlow serving infrastructure by triggering a null pointer dereference in the SavedModel loader's GetInitOp function. This is a denial-of-service risk for any internet-facing or internally-exposed TensorFlow inference endpoint. Patch immediately to TF 2.8.0, 2.7.1, 2.6.3, or 2.5.3; restrict serving endpoint access to trusted principals in the interim.
Is CVE-2022-23577 actively exploited?
Proof-of-concept exploit code is publicly available for CVE-2022-23577, increasing the risk of exploitation.
How to fix CVE-2022-23577?
1. Patch: Upgrade TensorFlow to 2.8.0, or apply cherry-pick backports to 2.7.1, 2.6.3, or 2.5.3. 2. Network hardening: Place TensorFlow serving endpoints behind authenticated API gateways; do not expose raw TF Serving ports to untrusted networks. 3. Process isolation: Run TF Serving workers in containers with auto-restart policies so crashes auto-recover (reduces blast radius while patching). 4. Detection: Alert on abnormal crash/restart rates in model serving containers; log null pointer/segfault signals from TF processes. 5. If running on unpatched versions, consider disabling dynamic SavedModel loading from untrusted sources.
What systems are affected by CVE-2022-23577?
This vulnerability affects the following AI/ML architecture patterns: model serving, training pipelines, inference.
What is the CVSS score for CVE-2022-23577?
CVE-2022-23577 has a CVSS v3.1 base score of 6.5 (MEDIUM). The EPSS exploitation probability is 0.79%.
What is the AI security impact?
Affected AI Architectures
MITRE ATLAS Techniques
AML.T0010.001AI SoftwareAML.T0029Denial of AI ServiceAML.T0049Exploit Public-Facing ApplicationCompliance Controls Affected
What are the technical details?
Original Advisory
Tensorflow is an Open Source Machine Learning Framework. The implementation of `GetInitOp` is vulnerable to a crash caused by dereferencing a null pointer. 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 adversary with low-privilege access (e.g., a data scientist account, a compromised CI/CD service account, or a legitimate but malicious user in a shared ML platform) submits a crafted SavedModel or triggers an inference request that causes GetInitOp to dereference a null pointer. The TensorFlow serving process crashes immediately with no error recovery. In a production environment without auto-restart, this takes the inference endpoint offline. An adversary could script repeated crash triggers to maintain a persistent denial-of-service condition against a competitor's shared ML inference cluster or a SaaS AI product.
Weaknesses (CWE)
CWE-476 — NULL Pointer Dereference: The product dereferences a pointer that it expects to be valid but is NULL.
- [Implementation] For any pointers that could have been modified or provided from a function that can return NULL, check the pointer for NULL before use. When working with a multithreaded or otherwise asynchronous environment, ensure that proper locking APIs are used to lock before the check, and unlock when it has finished [REF-1484].
- [Requirements] Select a programming language that is not susceptible to these issues.
Source: MITRE CWE corpus.
CVSS Vector
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:HReferences
- github.com/tensorflow/tensorflow/blob/a1320ec1eac186da1d03f033109191f715b2b130/tensorflow/cc/saved_model/loader_util.ccExploit3rd Party
- github.com/tensorflow/tensorflow/commit/4f38b1ac8e42727e18a2f0bde06d3bee8e77b250Patch3rd Party
- github.com/tensorflow/tensorflow/security/advisories/GHSA-8cxv-76p7-jxwrPatch3rd Party
Timeline
Related Vulnerabilities
CVE-2026-189489.9Feast: insecure UDF deserialization enables RCE
Same package: tensorflowCVE-2020-151969.9TensorFlow: heap OOB read in sparse/ragged count ops
Same package: tensorflowCVE-2020-152059.8TensorFlow: heap overflow in StringNGrams, ASLR bypass
Same package: tensorflowCVE-2019-167789.8TensorFlow: heap overflow in UnsortedSegmentSum op
Same package: tensorflowCVE-2020-152089.8TFLite: OOB read/write via tensor dimension mismatch
Same package: tensorflow