In TensorFlow before version 2.6.0 the implementation for tf.raw_ops.FractionalAvgPoolGrad can be tricked into accessing data outside of bounds of heap allocated buffers. The implementation does not validate that the input tensor is non-empty. Thus, code constructs an empty EigenDoubleMatrixMap and then accesses this buffer with indices that are outside of the empty area.
In TensorFlow before version 2.6.0 the implementation for tf.raw_ops.FractionalAvgPoolGrad can be tricked into accessing data outside of bounds of heap allocated buffers. The implementation does not validate that the input tensor is non-empty. Thus, code constructs an empty EigenDoubleMatrixMap and then accesses this buffer with indices that are outside of the empty area.
https://github.com/tensorflow/tensorflow/security/advisories/GHSA-hpv4-7p9c-mvfr https://github.com/tensorflow/tensorflow/commit/0f931751fb20f565c4e94aa6df58d54a003cdb30