CVE-2020-15193
Description
In Tensorflow before versions 2.2.1 and 2.3.1, 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. However, there is nothing stopping users from passing in a Python object instead of a tensor. The uninitialized memory address is due to a `reinterpret_cast` 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.
How to fix
- tensorflow2.2.1, 2.3.1 or later
- tensorflow-cpu2.2.1, 2.3.1 or later
- tensorflow-gpu2.2.1, 2.3.1 or later
Affected products
| Product | Affected versions | Fixed in |
|---|---|---|
| google tensorflow | 2.2.0, 2.3.0 | — |
| opensuse leap | 15.2 | — |
| tensorflowPyPI | 2.2.0 ≤ v < 2.2.1, 2.3.0 ≤ v < 2.3.1, v < 2.2.1, 2.3.0rc0 ≤ v < 2.3.1 | 2.2.1, 2.3.1 |
| tensorflow-cpuPyPI | 2.2.0 ≤ v < 2.2.1, 2.3.0 ≤ v < 2.3.1, v < 2.2.1, 2.3.0rc0 ≤ v < 2.3.1 | 2.2.1, 2.3.1 |
| tensorflow-gpuPyPI | 2.2.0 ≤ v < 2.2.1, 2.3.0 ≤ v < 2.3.1, v < 2.2.1, 2.3.0rc0 ≤ v < 2.3.1 | 2.2.1, 2.3.1 |
Check your version
Choose the product and enter the version you run to see whether it falls in the affected range.
Attack Surface Management
See the assets tied to your domains, the technologies they run and known vulnerabilities in one list.