1## TFSA-2022-035: Crash when type cannot be specialized 2 3### CVE Number 4CVE-2022-23572 5 6### Impact 7Under certain scenarios, TensorFlow can fail to specialize a type during [shape inference](https://github.com/tensorflow/tensorflow/blob/a1320ec1eac186da1d03f033109191f715b2b130/tensorflow/core/framework/shape_inference.cc#L168-L174): 8 9```cc 10void InferenceContext::PreInputInit( 11 const OpDef& op_def, const std::vector<const Tensor*>& input_tensors, 12 const std::vector<ShapeHandle>& input_tensors_as_shapes) { 13 const auto ret = full_type::SpecializeType(attrs_, op_def); 14 DCHECK(ret.status().ok()) << "while instantiating types: " << ret.status(); 15 ret_types_ = ret.ValueOrDie(); 16 // ... 17} 18``` 19 20However, `DCHECK` is a no-op in production builds and an assertion failure in debug builds. In the first case execution proceeds to the `ValueOrDie` line. This results in an assertion failure as `ret` contains an error `Status`, not a value. In the second case we also get a crash due to the assertion failure. 21 22### Patches 23We have patched the issue in GitHub commit [cb164786dc891ea11d3a900e90367c339305dc7b](https://github.com/tensorflow/tensorflow/commit/cb164786dc891ea11d3a900e90367c339305dc7b). 24 25The fix will be included in TensorFlow 2.8.0. We will also cherrypick this commit on TensorFlow 2.7.1, and TensorFlow 2.6.3, as these are also affected and still in supported range. 26 27### For more information 28Please 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. 29