Bad alloc in `StringNGrams` caused by integer conversion
Medium5.5CVE-2021-37646 · Published Aug 25, 2021 · updated Jul 8, 2026
### Impact The implementation of `tf.raw_ops.StringNGrams` is vulnerable to an integer overflow issue caused by converting a signed integer value to an unsigned one and then allocating memory based on this value. ```python import tensorflow as tf tf.raw_ops.StringNGrams( data=['',''], data_splits=[0,2], separator=' '*100, ngram_widths=[-80,0,0,-60], left_pad=' ', right_pad=' ', pad_width=100, preserve_short_sequences=False) ``` The [implementation](https://github.com/tensorflow/tensorflow/blob/8d72537c6abf5a44103b57b9c2e22c14f5f49698/tensorflow/core/kernels/string_ngrams_op.cc#L184) calls `reserve` on a `tstring` with a value that sometimes can be negative if user supplies negative `ngram_widths`. The `reserve` method calls `TF_TString_Reserve` which has an `unsigned long` argument for the size of the buffer. Hence, the implicit conversion transforms the negative value to a large integer. ### Patches We have patched the issue in GitHub commit [c283e542a3f422420cfdb332414543b62fc4e4a5](https://github.com/tensorflow/tensorflow/commit/c283e542a3f422420cfdb332414543b62fc4e4a5). The fix will be included in TensorFlow 2.6.0. We will also cherrypick this commit on Ten...
Affected versions
| Package | Affected | Fixed in |
|---|---|---|
| tensorflow PyPI | < 2.3.4 | 2.3.4 |
| >= 2.4.0, < 2.4.3 | 2.4.3 | |
| >= 2.5.0, < 2.5.1 | 2.5.1 |
Details and references
### Impact The implementation of `tf.raw_ops.StringNGrams` is vulnerable to an integer overflow issue caused by converting a signed integer value to an unsigned one and then allocating memory based on this value. ```python import tensorflow as tf tf.raw_ops.StringNGrams( data=['',''], data_splits=[0,2], separator=' '*100, ngram_widths=[-80,0,0,-60], left_pad=' ', right_pad=' ', pad_width=100, preserve_short_sequences=False) ``` The [implementation](https://github.com/tensorflow/tensorflow/blob/8d72537c6abf5a44103b57b9c2e22c14f5f49698/tensorflow/core/kernels/string_ngrams_op.cc#L184) calls `reserve` on a `tstring` with a value that sometimes can be negative if user supplies negative `ngram_widths`. The `reserve` method calls `TF_TString_Reserve` which has an `unsigned long` argument for the size of the buffer. Hence, the implicit conversion transforms the negative value to a large integer. ### Patches We have patched the issue in GitHub commit [c283e542a3f422420cfdb332414543b62fc4e4a5](https://github.com/tensorflow/tensorflow/commit/c283e542a3f422420cfdb332414543b62fc4e4a5). 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](https://github.com/tensorflow/tensorflow/blob/master/SECURITY.md) 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.
- CVSS 3.1
- CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H
- Severity from
- GitHub (reviewed advisory)
- Weakness
- CWE-681
- Also known as
- BIT-tensorflow-2021-37646, CVE-2021-37646, PYSEC-2021-268, PYSEC-2021-559, PYSEC-2021-757
- github.com/tensorflow/tensorflow/security/advisories/GHSA-h6jh-7gv5-28vg
- nvd.nist.gov/vuln/detail/CVE-2021-37646
- github.com/tensorflow/tensorflow/commit/c283e542a3f422420cfdb332414543b62fc4e4a5
- github.com/pypa/advisory-database/tree/main/vulns/tensorflow-cpu/PYSEC-2021-559.yaml
- github.com/pypa/advisory-database/tree/main/vulns/tensorflow-gpu/PYSEC-2021-757.yaml
- github.com/pypa/advisory-database/tree/main/vulns/tensorflow/PYSEC-2021-268.yaml
- github.com/tensorflow/tensorflow
More TensorFlow advisories
All TensorFlow| Date | Advisory | Severity | Fixed in |
|---|---|---|---|
| Aug 252021 | Heap out of bounds access in sparse reduction operations High7.3Aug 25, 2021 | High7.3 | 2.3.4+2 more |
| Aug 252021 | Floating point exception in `SparseDenseCwiseDiv`Medium5.5Aug 25, 2021 | Medium5.5 | 2.3.4+2 more |
| Aug 252021 | Null pointer dereference in `CompressElement`High7.7Aug 25, 2021 | High7.7 | 2.3.4+2 more |
| Aug 252021 | Null pointer dereference in `RaggedTensorToTensor`High7.7Aug 25, 2021 | High7.7 | 2.3.4+2 more |
| Aug 252021 | Null pointer dereference and heap OOB read in operations restoring tensorsHigh8.4Aug 25, 2021 | High8.4 | 2.3.4+2 more |
| Aug 252021 | Integer division by 0 in sparse reshapingMedium5.5Aug 25, 2021 | Medium5.5 | 2.3.4+2 more |