CVE-2021-41202

TensorFlow is an open source platform for machine learning. In affected versions while calculating the size of the output within the `tf.range` kernel, there is a conditional statement of type `int64 = condition ? int64 : double`. Due to C++ implicit conversion rules, both branches of the condition will be cast to `double` and the result would be truncated before the assignment. This result in overflows. 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.

TensorFlow is an open source platform for machine learning. In affected versions while calculating the size of the output within the `tf.range` kernel, there is a conditional statement of type `int64 = condition ? int64 : double`. Due to C++ implicit conversion rules, both branches of the condition will be cast to `double` and the result would be truncated before the assignment. This result in overflows. 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.

TensorFlow is an open source platform for machine learning. In affected versions while calculating the size of the output within the `tf.range` kernel, there is a conditional statement of type `int64 = condition ? int64 : double`. Due to C++ implicit conversion rules, both branches of the condition will be cast to `double` and the result would be truncated before the assignment. This result in overflows. 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.

### Impact While calculating the size of the output within the `tf.range` kernel, there is a conditional statement of type `int64 = condition ? int64 : double`. Due to C++ implicit conversion rules, both branches of the condition will be cast to `double` and the result would be truncated before the assignment. This result in overflows: ```python import tensorflow as tf tf.sparse.eye(num_rows=9223372036854775807, num_columns=None) ``` Similarly, `tf.range` would result in crashes due to overflows if the start or end point are too large. ```python import tensorflow as tf tf.range(start=-1e+38, limit=1) ``` ### Patches We have patched the issue in GitHub commits [6d94002a09711d297dbba90390d5482b76113899](https://github.com/tensorflow/tensorflow/commit/6d94002a09711d297dbba90390d5482b76113899) (merging [#51359](https://github.com/tensorflow/tensorflow/pull/51359)) and [1b0e0ec27e7895b9985076eab32445026ae5ca94](https://github.com/tensorflow/tensorflow/commit/1b0e0ec27e7895b9985076eab32445026ae5ca94) (merging [#51711](https://github.com/tensorflow/tensorflow/pull/51711)). 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 [GitHub issue](https://github.com/tensorflow/tensorflow/issues/46912), [GitHub issue](https://github.com/tensorflow/tensorflow/issues/46899) and [GitHub issue](https://github.com/tensorflow/tensorflow/issues/46889).

TensorFlow es una plataforma de código abierto para el aprendizaje automático. En las versiones afectadas, mientras se calcula el tamaño de la salida dentro del núcleo "tf.range", se presenta una sentencia condicional de tipo "int64 = condición ? int64 : double". Debido a las reglas de conversión implícitas de C++, ambas ramas de la condición se convertirán en "double" y el resultado se truncará antes de la asignación. Esto resulta en desbordamientos. La corrección será incluida en TensorFlow versión 2.7.0. También será incluida este commit en TensorFlow versión 2.6.1, TensorFlow versión 2.5.2, y TensorFlow versión 2.4.4, ya que estos también están afectados y todavía están en el rango admitido