CVE-2020-15193: Memory corruption in Tensorflow
Impact The implementation of dlpack.todlpack 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/tfewrapper.cc#L1361
However, there is nothing stopping users from passing in a Python object instead of a tensor. python In [2]: tf.experimental.dlpack.todlpack([2]) ==1720623==WARNING: MemorySanitizer: use-of-uninitialized-value #0 0x55b0ba5c410a in tensorflow::(anonymous namespace)::GetTensorFromHandle(TFETensorHandle, TFStatus) thirdparty/tensorflow/c/eager/dlpack.cc:46:7 #1 0x55b0ba5c38f4 in tensorflow::TFEHandleToDLPack(TFETensorHandle, TFStatus) thirdparty/tensorflow/c/eager/dlpack.cc:252:26 ...
The uninitialized memory address is due to a reinterpretcast https://github.com/tensorflow/tensorflow/blob/0e68f4d3295eb0281a517c3662f6698992b7b2cf/tensorflow/python/eager/pywraptensor.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.
Other sources
In Tensorflow before versions 2.2.1 and 2.3.1, the implementation of dlpack.todlpack 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. However, there is nothing stopping users from passing in a Python object instead of a tensor. The uninitialized memory address is due to a reinterpretcast Since the PyObject is a Python object, not a TensorFlow Tensor, the cast to EagerTensor fails. The issue is patched in commit 22e07fb204386768e5bcbea563641ea11f96ceb8 and is released in TensorFlow versions 2.2.1, or 2.3.1.
Affected Software
Remediation
Event History
Frequently Asked Questions
What is the severity of CVE-2020-15193?
CVE-2020-15193 has been classified as a high severity vulnerability due to the potential for memory corruption.
How do I fix CVE-2020-15193?
To fix CVE-2020-15193, upgrade the affected TensorFlow packages to version 2.3.1 or later.
Which versions of TensorFlow are affected by CVE-2020-15193?
CVE-2020-15193 affects TensorFlow versions 2.2.0 and 2.3.0.
What could happen if CVE-2020-15193 is exploited?
If exploited, CVE-2020-15193 may lead to memory corruption, potentially resulting in application crashes or arbitrary code execution.
Is CVE-2020-15193 relevant to both TensorFlow CPU and GPU?
Yes, CVE-2020-15193 affects both TensorFlow GPU and TensorFlow CPU implementations.