Advisoriesby IbisCrash in `tf.math.segment_*` operations
Impact
The implementation of tf.math.segment_* operations results in a CHECK-fail related abort (and denial of service) if a segment id in segment_ids is large.
```python
import tensorflow as tf
tf.math.segment_max(data=np.ones((1,10,1)), segment_ids=[1676240524292489355])
tf.math.segment_min(data=np.ones((1,10,1)), segment_ids=[1676240524292489355])
tf.math.segment_mean(data=np.ones((1,10,1)), segment_ids=[1676240524292489355])
tf.math.segment_sum(data=np.ones((1,10,1)), segment_ids=[1676240524292489355])
tf.math.segment_prod(data=np.ones((1,10,1)), segment_ids=[1676240524292489355])
`
This is similar to [CVE-2021-29584](https://github.com/tensorflow/tensorflow/blob/3a74f0307236fe206b046689c4d76f57c9b74eee/tensorflow/security/advisory/tfsa-2021-071.md) (and similar other reported vulnerabilities in TensorFlow, localized to specific APIs): the [implementation](https://github.com/tensorflow/tensorflow/blob/dae66e518c88de9c11718cd0f8f40a0b666a90a0/tensorflow/core/kernels/segment_reduction_ops_impl.h) (both on CPU and GPU) computes the output shape using [AddDim](https://github.com/tensorflow/tensorflow/blob/0b6b491d21d6a4eb5fbab1cca565bc1e94ca9543/tensorflow/core/framework/tensor_shape.cc#L395-L408). However, if the number of elements in the tensor overflows an int64_t value, AddDim results in a CHECK failure which provokes a std::abort. Instead, code...
Affected versions
| Package | Affected | Fixed in |
|---|---|---|
| tensorflow PyPI | >= 2.6.0, < 2.6.1 | 2.6.1 |
| >= 2.5.0, < 2.5.2 | 2.5.2 | |
| < 2.4.4 | 2.4.4 |
Details and references
### Impact The implementation of `tf.math.segment_*` operations results in a `CHECK`-fail related abort (and denial of service) if a segment id in `segment_ids` is large. ```python import tensorflow as tf tf.math.segment_max(data=np.ones((1,10,1)), segment_ids=[1676240524292489355]) tf.math.segment_min(data=np.ones((1,10,1)), segment_ids=[1676240524292489355]) tf.math.segment_mean(data=np.ones((1,10,1)), segment_ids=[1676240524292489355]) tf.math.segment_sum(data=np.ones((1,10,1)), segment_ids=[1676240524292489355]) tf.math.segment_prod(data=np.ones((1,10,1)), segment_ids=[1676240524292489355]) ``` This is similar to [CVE-2021-29584](https://github.com/tensorflow/tensorflow/blob/3a74f0307236fe206b046689c4d76f57c9b74eee/tensorflow/security/advisory/tfsa-2021-071.md) (and similar other reported vulnerabilities in TensorFlow, localized to specific APIs): the [implementation](https://github.com/tensorflow/tensorflow/blob/dae66e518c88de9c11718cd0f8f40a0b666a90a0/tensorflow/core/kernels/segment_reduction_ops_impl.h) (both on CPU and GPU) computes the output shape using [`AddDim`](https://github.com/tensorflow/tensorflow/blob/0b6b491d21d6a4eb5fbab1cca565bc1e94ca9543/tensorflow/core/framework/tensor_shape.cc#L395-L408). However, if the number of elements in the tensor overflows an `int64_t` value, `AddDim` results in a `CHECK` failure which provokes a `std::abort`. Instead, code should use [`AddDimWithStatus`](https://github.com/tensorflow/tensorflow/blob/0b6b491d21d6a4eb5fbab1cca565bc1e94ca9543/tensorflow/core/framework/tensor_shape.cc#L410-L440). ### Patches We have patched the issue in GitHub commit [e9c81c1e1a9cd8dd31f4e83676cab61b60658429](https://github.com/tensorflow/tensorflow/commit/e9c81c1e1a9cd8dd31f4e83676cab61b60658429) (merging [#51733](https://github.com/tensorflow/tensorflow/pull/51733)). The fix will be included in TensorFlow 2.7.0. We will also cherrypick this commit on TensorFlow 2.6.1, TensorFlow 2.5.2, and TensorFlow 2.4.4, 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 externally via a [GitHub issue](https://github.com/tensorflow/tensorflow/issues/46888).
- 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-190
- Also known as
- BIT-tensorflow-2021-41195, CVE-2021-41195, PYSEC-2021-842, PYSEC-2021-844, PYSEC-2021-846
- github.com/tensorflow/tensorflow/security/advisories/GHSA-cq76-mxrc-vchh
- nvd.nist.gov/vuln/detail/CVE-2021-41195
- github.com/tensorflow/tensorflow/issues/46888
- github.com/tensorflow/tensorflow/pull/51733
- github.com/tensorflow/tensorflow/commit/e9c81c1e1a9cd8dd31f4e83676cab61b60658429
- github.com/pypa/advisory-database/tree/main/vulns/tensorflow-cpu/PYSEC-2021-844.yaml
- github.com/pypa/advisory-database/tree/main/vulns/tensorflow-gpu/PYSEC-2021-846.yaml
- github.com/pypa/advisory-database/tree/main/vulns/tensorflow/PYSEC-2021-842.yaml
- github.com/tensorflow/tensorflow
More TensorFlow advisories
All TensorFlow
Incomplete validation in boosted trees codeCVE-2021-41208fixed in 2.4.4 +2Crash in `max_pool3d` when size argument is 0 or negativeCVE-2021-41196fixed in 2.4.4 +2Crashes due to overflow and `CHECK`-fail in ops with large tensor shapesCVE-2021-41197fixed in 2.4.4 +2Overflow/crash in `tf.tile` when tiling tensor is largeCVE-2021-41198fixed in 2.4.4 +2Overflow/crash in `tf.image.resize` when size is largeCVE-2021-41199fixed in 2.4.4 +2Incomplete validation in `tf.summary.create_file_writer`CVE-2021-41200fixed in 2.4.4 +2