Memory corruption in Tensorflow

Overview

Impact

The implementation of dlpack.to_dlpack can be made to use uninitialized memory resulting in further memory corruption. This is because the pybind11 glue code assumes that the argument is a tensor: https://github.com/tensorflow/tensorflow/blob/0e68f4d3295eb0281a517c3662f6698992b7b2cf/tensorflow/python/tfe_wrapper.cc#L1361

However, there is nothing stopping users from passing in a Python object instead of a tensor.

In [2]: tf.experimental.dlpack.to_dlpack([2])                                                                                                                                            
==1720623==WARNING: MemorySanitizer: use-of-uninitialized-value                                                                                                                            
    #0 0x55b0ba5c410a in tensorflow::(anonymous namespace)::GetTensorFromHandle(TFE_TensorHandle*, TF_Status*) third_party/tensorflow/c/eager/dlpack.cc:46:7
    #1 0x55b0ba5c38f4 in tensorflow::TFE_HandleToDLPack(TFE_TensorHandle*, TF_Status*) third_party/tensorflow/c/eager/dlpack.cc:252:26
... 

The uninitialized memory address is due to a reinterpret_cast https://github.com/tensorflow/tensorflow/blob/0e68f4d3295eb0281a517c3662f6698992b7b2cf/tensorflow/python/eager/pywrap_tensor.cc#L848-L850

Since the PyObject is a Python object, not a TensorFlow Tensor, the cast to EagerTensor fails.

Patches

We have patched the issue in 22e07fb204386768e5bcbea563641ea11f96ceb8 and will release a patch release for all affected versions.

We recommend users to upgrade to TensorFlow 2.2.1 or 2.3.1.

For more information

Please consult our security guide for more information regarding the security model and how to contact us with issues and questions.

Attribution

This vulnerability has been reported by members of the Aivul Team from Qihoo 360.

Affected packages

• tensorflow >= 2.2.0, < 2.2.1
• tensorflow >= 2.3.0, < 2.3.1
• tensorflow-cpu >= 2.2.0, < 2.2.1
• tensorflow-cpu >= 2.3.0, < 2.3.1
• tensorflow-gpu >= 2.2.0, < 2.2.1
• tensorflow-gpu >= 2.3.0, < 2.3.1

Remediation

Upgrade to a patched release:

• tensorflow 2.2.1
• tensorflow 2.3.1
• tensorflow-cpu 2.2.1
• tensorflow-cpu 2.3.1
• tensorflow-gpu 2.2.1
• tensorflow-gpu 2.3.1