CVE-2021-37672 is a medium-severity (CVSS 5.5) Out-of-bounds Read vulnerability in tensorflow. A fix is available for tensorflow — see the affected versions and patch details below.
Heap OOB in `SdcaOptimizerV2`
EPSS Exploitation Probability
Probability of exploitation in the next 30 days, from FIRST.org EPSS.
How urgent is this, really
CVE-2021-37672 by exploitation likelihood (EPSS) against impact (CVSS). Outside the shaded patch-first corner.
Where this sits among everything scored
Of 385,738 CVEs with a current EPSS score, this one falls in the < 10% band (highlighted). Counts from FIRST.org, log-scaled.
Real-World Exposure
tensorflow🐍tensorflow🐍tensorflow🐍tensorflow-cpu🐍tensorflow-cpu🐍tensorflow-cpu🐍tensorflow-gpu🐍tensorflow-gpu+1 moreReal-time download stats are indexed for npm and PyPI packages. This vulnerability affects PyPI packages — download data is not available via public APIs for these ecosystems.
Description
Impact
An attacker can read from outside of bounds of heap allocated data by sending specially crafted illegal arguments to tf.raw_ops.SdcaOptimizerV2:
import tensorflow as tf
tf.raw_ops.SdcaOptimizerV2(
sparse_example_indices=[[1]],
sparse_feature_indices=[[1]],
sparse_feature_values=[[1.0,2.0]],
dense_features=[[1.0]],
example_weights=[1.0],
example_labels=[],
sparse_indices=[1],
sparse_weights=[1.0],
dense_weights=[[1.0]],
example_state_data=[[100.0,100.0,100.0,100.0]],
loss_type='logistic_loss',
l1=100.0,
l2=100.0,
num_loss_partitions=1,
num_inner_iterations=1,
adaptive=True)
The implementation does not check that the length of example_labels is the same as the number of examples.
Patches
We have patched the issue in GitHub commit a4e138660270e7599793fa438cd7b2fc2ce215a6.
The fix will be included in TensorFlow 2.6.0. We will also cherrypick this commit on TensorFlow 2.5.1, TensorFlow 2.4.3, and TensorFlow 2.3.4, as these are also affected and still in supported range.
For more information
Please consult our security guide for more information regarding the security model and how to contact us with issues and questions.
Attribution
This vulnerability has been reported by members of the Aivul Team from Qihoo 360.
Affected Packages
| Ecosystem | Package | Vulnerable range | Fix |
|---|---|---|---|
| 🐍PyPI | tensorflow | all versions | 2.3.4pip install --upgrade 'tensorflow==2.3.4' |
| 🐍PyPI | tensorflow | ≥ 2.4.0&&< 2.4.3 | 2.4.3pip install --upgrade 'tensorflow==2.4.3' |
| 🐍PyPI | tensorflow | ≥ 2.5.0&&< 2.5.1 | 2.5.1pip install --upgrade 'tensorflow==2.5.1' |
| 🐍PyPI | tensorflow-cpu | all versions | 2.3.4pip install --upgrade 'tensorflow-cpu==2.3.4' |
| 🐍PyPI | tensorflow-cpu | ≥ 2.4.0&&< 2.4.3 | 2.4.3pip install --upgrade 'tensorflow-cpu==2.4.3' |
| 🐍PyPI | tensorflow-cpu | ≥ 2.5.0&&< 2.5.1 | 2.5.1pip install --upgrade 'tensorflow-cpu==2.5.1' |
Affected Products
tensorflowgoogleDetection & mitigation playbook
Open-source dependencyDetect
Scan your dependency tree (package-lock.json, pnpm-lock.yaml, requirements.txt, go.sum, etc.) for tensorflow, including transitive dependencies — a direct dependency you never call can still pull in a vulnerable version.
Fix
Update tensorflow to 2.3.4 or later, then make sure no transitive (indirect) dependency still pins the vulnerable range — O3 confirms CVE-2021-37672 is resolved across your whole dependency graph.
Workarounds
Stop feeding it untrusted input: reject or quarantine files and payloads from unverified sources until you can upgrade, restrict accepted formats to the ones you actually need, and run the parsing or decoding step in a least-privileged sandbox or short-lived worker so a crash or corrupted read cannot reach the rest of the process.
Frequently Asked Questions
Is CVE-2021-37672 in your dependencies?
Find it across PyPI, including transitive dependencies.