1## TFSA-2020-019: Crash due to invalid splits in SparseCountSparseOutput 2 3### CVE Number 4CVE-2020-15197 5 6### Impact 7The `SparseCountSparseOutput` implementation does not validate that the input 8arguments form a valid sparse tensor. In particular, there is no validation that 9the `indices` tensor has rank 2. This tensor must be a matrix because code 10assumes its elements are [accessed as elements of a 11matrix](https://github.com/tensorflow/tensorflow/blob/0e68f4d3295eb0281a517c3662f6698992b7b2cf/tensorflow/core/kernels/count_ops.cc#L185): 12```cc 13 const auto indices_values = indices.matrix<int64>(); 14``` 15 16However, malicious users can pass in tensors of different rank, resulting in a 17`CHECK` assertion failure and a crash. This can be used to cause denial of 18service in serving installations, if users are allowed to control the components 19of the input sparse tensor. 20 21### Vulnerable Versions 22TensorFlow 2.3.0. 23 24### Patches 25We have patched the issue in 26[3cbb917b4714766030b28eba9fb41bb97ce9ee02](https://github.com/tensorflow/tensorflow/commit/3cbb917b4714766030b28eba9fb41bb97ce9ee02) 27and will release a patch release. 28 29We recommend users to upgrade to TensorFlow 2.3.1. 30 31### For more information 32Please consult [our security 33guide](https://github.com/tensorflow/tensorflow/blob/master/SECURITY.md) for 34more information regarding the security model and how to contact us with issues 35and questions. 36 37### Attribution 38This vulnerability has been discovered through a variant analysis of [a 39vulnerability reported by members of the Aivul Team from Qihoo 40360](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/security/advisory/tfsa-2020-015.md). 41