CVE-2021-37692 Segfault on strings tensors with mistmatched dimensions in TensorFlow
An attacker can cause a segmentation fault and denial of service in TensorFlow by providing string tensors with mismatched dimensions that fail encoding.
- Exploited in the wild
- Not known
- Login needed
- Any signed-in user
- Attackable
- Local or same network only
- Victim action needed
- Not needed
- Fix
- Available
What is CVE-2021-37692?
CVE-2021-37692 is a medium-severity vulnerability in Google Tensorflow, published on 12 Aug 2021. It can only be exploited from the same machine or local network, with no action from the victim.
Official description
TensorFlow is an end-to-end open source platform for machine learning. In affected versions under certain conditions, Go code can trigger a segfault in string deallocation. For string tensors, `C.TF_TString_Dealloc` is called during garbage collection within a finalizer function. However, tensor structure isn't checked until encoding to avoid a performance penalty. The current method for dealloc assumes that encoding succeeded, but segfaults when a string tensor is garbage collected whose encoding failed (e.g., due to mismatched dimensions). To fix this, the call to set the finalizer function is deferred until `NewTensor` returns and, if encoding failed for a string tensor, deallocs are determined based on bytes written. We have patched the issue in GitHub commit 8721ba96e5760c229217b594f6d2ba332beedf22. The fix will be included in TensorFlow 2.6.0. We will also cherrypick this commit on TensorFlow 2.5.1, which is the other affected version.
Weakness: CWE-20
Which versions are affected by CVE-2021-37692?
10 open source packages ship an affected version, including tensorflow, tensorflow-cpu and tensorflow-gpu. It affects 1 product.
| Package | Affected | Fixed in |
|---|---|---|
tensorflow Bitnami | From 2.5.0 | 2.6.0 |
tensorflow PyPI | From 2.3.0 | 2.3.4, 2.4.3 |
tensorflow PyPI | From 2.5.0rc0 | 2.5.1 |
tensorflow PyPI | From 2.5.0rc0 | 2.5.1 |
tensorflow-cpu PyPI | From 2.3.0 | 2.3.4, 2.4.3 |
tensorflow-cpu PyPI | From 2.5.0rc0 | 2.5.1 |
4 more packages
tensorflow-cpu PyPI | From 2.5.0rc0 | 2.5.1 |
tensorflow-gpu PyPI | From 2.3.0 | 2.3.4, 2.4.3 |
tensorflow-gpu PyPI | From 2.5.0rc0 | 2.5.1 |
tensorflow-gpu PyPI | From 2.5.0rc0 | 2.5.1 |
Products include Google Tensorflow.
How do I fix CVE-2021-37692?
Upgrade tensorflow to a fixed version for your release line: 2.6.0, 2.3.4, 2.4.3 or 2.5.1.
Is CVE-2021-37692 being exploited?
There is no known exploitation of CVE-2021-37692 in the wild and no public exploit code.
Other vulnerabilities in Google Tensorflow
- CVE-2023-25668 TensorFlow vulnerable to heap out-of-buffer read in the QuantizeAndDequantize operation
- CVE-2018-7575 Integer Overflow or Wraparound in Google TensorFlow
- CVE-2021-37678 Arbitrary code execution due to YAML deserialization
- CVE-2021-35958 Path traversal in n/a
- CVE-2020-15202 Integer truncation in Shard API usage
Sources
Also known as BIT-tensorflow-2021-37692, GHSA-cmgw-8vpc-rc59, PYSEC-2021-314, PYSEC-2021-605, PYSEC-2021-803.