Impact
The simplifyBroadcast
function in the MLIR-TFRT infrastructure in TensorFlow is vulnerable to a segfault (hence, denial of service), if called with scalar shapes.
size_t maxRank = 0;
for (auto shape : llvm::enumerate(shapes)) {
auto found_shape = analysis.dimensionsForShapeTensor(shape.value());
if (!found_shape) return {};
shapes_found.push_back(*found_shape);
maxRank = std::max(maxRank, found_shape->size());
}
SmallVector<const ShapeComponentAnalysis::SymbolicDimension*>
joined_dimensions(maxRank);
If all shapes are scalar, then maxRank
is 0, so we build an empty SmallVector
.
Patches
We have patched the issue in GitHub commit 35f0fabb4c178253a964d7aabdbb15c6a398b69a.
The fix will be included in TensorFlow 2.8.0. This is the only affected version.
For more information
Please consult our security guide for more information regarding the security model and how to contact us with issues and questions.
References
- https://github.com/tensorflow/tensorflow/security/advisories/GHSA-gwcx-jrx4-92w2
- https://github.com/tensorflow/tensorflow/commit/35f0fabb4c178253a964d7aabdbb15c6a398b69a
- https://github.com/tensorflow/tensorflow/blob/274df9b02330b790aa8de1cee164b70f72b9b244/tensorflow/compiler/mlir/tfrt/jit/transforms/tf_cpurt_symbolic_shape_optimization.cc#L149-L205
- https://nvd.nist.gov/vuln/detail/CVE-2022-23593
- https://github.com/advisories/GHSA-gwcx-jrx4-92w2