First published: Fri May 14 2021(Updated: )
### Impact Calling `tf.raw_ops.RaggedTensorToVariant` with arguments specifying an invalid ragged tensor results in a null pointer dereference: ```python import tensorflow as tf input_tensor = tf.constant([], shape=[0, 0, 0, 0, 0], dtype=tf.float32) filter_tensor = tf.constant([], shape=[0, 0, 0, 0, 0], dtype=tf.float32) tf.raw_ops.Conv3D(input=input_tensor, filter=filter_tensor, strides=[1, 56, 56, 56, 1], padding='VALID', data_format='NDHWC', dilations=[1, 1, 1, 23, 1]) ``` ```python import tensorflow as tf input_tensor = tf.constant([], shape=[2, 2, 2, 2, 0], dtype=tf.float32) filter_tensor = tf.constant([], shape=[0, 0, 2, 6, 2], dtype=tf.float32) tf.raw_ops.Conv3D(input=input_tensor, filter=filter_tensor, strides=[1, 56, 39, 34, 1], padding='VALID', data_format='NDHWC', dilations=[1, 1, 1, 1, 1]) ``` The implementation of [`RaggedTensorToVariant` operations](https://github.com/tensorflow/tensorflow/blob/904b3926ed1c6c70380d5313d282d248a776baa1/tensorflow/core/kernels/ragged_tensor_to_variant_op.cc#L39-L40) does not validate that the ragged tensor argument is non-empty: ```cc int ragged_rank = batched_ragged.ragged_rank(); auto batched_splits_top_vec = batched_ragged.splits(0).vec<SPLIT_TYPE>(); ``` Since `batched_ragged` contains no elements, `batched_ragged.splits` is a null vector, thus `batched_ragged.splits(0)` will result in dereferencing `nullptr`. ### Patches We have patched the issue in GitHub commit [b055b9c474cd376259dde8779908f9eeaf097d93](https://github.com/tensorflow/tensorflow/commit/b055b9c474cd376259dde8779908f9eeaf097d93). The fix will be included in TensorFlow 2.5.0. We will also cherrypick this commit on TensorFlow 2.4.2, TensorFlow 2.3.3, TensorFlow 2.2.3 and TensorFlow 2.1.4, 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 Yakun Zhang and Ying Wang of Baidu X-Team.
Credit: security-advisories@github.com security-advisories@github.com
Affected Software | Affected Version | How to fix |
---|---|---|
Google TensorFlow | <2.1.4 | |
Google TensorFlow | >=2.2.0<2.2.3 | |
Google TensorFlow | >=2.3.0<2.3.3 | |
Google TensorFlow | >=2.4.0<2.4.2 | |
pip/tensorflow-gpu | >=2.4.0<2.4.2 | 2.4.2 |
pip/tensorflow-gpu | >=2.3.0<2.3.3 | 2.3.3 |
pip/tensorflow-gpu | >=2.2.0<2.2.3 | 2.2.3 |
pip/tensorflow-gpu | <2.1.4 | 2.1.4 |
pip/tensorflow-cpu | >=2.4.0<2.4.2 | 2.4.2 |
pip/tensorflow-cpu | >=2.3.0<2.3.3 | 2.3.3 |
pip/tensorflow-cpu | >=2.2.0<2.2.3 | 2.2.3 |
pip/tensorflow-cpu | <2.1.4 | 2.1.4 |
pip/tensorflow | >=2.4.0<2.4.2 | 2.4.2 |
pip/tensorflow | >=2.3.0<2.3.3 | 2.3.3 |
pip/tensorflow | >=2.2.0<2.2.3 | 2.2.3 |
pip/tensorflow | <2.1.4 | 2.1.4 |
Sign up to SecAlerts for real-time vulnerability data matched to your software, aggregated from hundreds of sources.
CVE-2021-29516 is classified as a medium severity vulnerability due to its potential to cause application crashes.
To fix CVE-2021-29516, upgrade TensorFlow to version 2.4.2 or above.
CVE-2021-29516 affects TensorFlow versions prior to 2.1.4, 2.2.0 to 2.2.3, 2.3.0 to 2.3.3, and 2.4.0 to 2.4.2.
CVE-2021-29516 is a null pointer dereference vulnerability that can occur during the execution of specific TensorFlow operations.
CVE-2021-29516 is generally considered less likely to be remotely exploitable, as it typically requires local execution of vulnerable TensorFlow code.