OOB read in `MatrixTriangularSolve`

Impact

The implementation of MatrixTriangularSolve fails to terminate kernel execution if one validation condition fails:

void ValidateInputTensors(OpKernelContext* ctx, const Tensor& in0,
                            const Tensor& in1) override {
  OP_REQUIRES(
      ctx, in0.dims() >= 2,
      errors::InvalidArgument("In[0] ndims must be >= 2: ", in0.dims()));

  OP_REQUIRES(
      ctx, in1.dims() >= 2,
      errors::InvalidArgument("In[0] ndims must be >= 2: ", in1.dims()));
}
  
void Compute(OpKernelContext* ctx) override {
  const Tensor& in0 = ctx->input(0);
  const Tensor& in1 = ctx->input(1);

  ValidateInputTensors(ctx, in0, in1);

  MatMulBCast bcast(in0.shape().dim_sizes(), in1.shape().dim_sizes());
  ...
}

Since OP_REQUIRES only sets ctx->status() to a non-OK value and calls return, this allows malicious attackers to trigger an out of bounds read:

import tensorflow as tf
import numpy as np

matrix_array = np.array([])
matrix_tensor = tf.convert_to_tensor(np.reshape(matrix_array,(1,0)),dtype=tf.float32)
rhs_array = np.array([])
rhs_tensor = tf.convert_to_tensor(np.reshape(rhs_array,(0,1)),dtype=tf.float32)

tf.raw_ops.MatrixTriangularSolve(matrix=matrix_tensor,rhs=rhs_tensor,lower=False,adjoint=False)

As the two input tensors are empty, the OP_REQUIRES in ValidateInputTensors should fire and interrupt execution. However, given the implementation of OP_REQUIRES, after the in0.dims() >= 2 fails, execution moves to the initialization of the bcast object. This initialization is done with invalid data and results in heap OOB read.

Patches

We have patched the issue in GitHub commit 480641e3599775a8895254ffbc0fc45621334f68.

The fix will be included in TensorFlow 2.5.0. We will also cherrypick this commit on TensorFlow 2.4.2, TensorFlow 2.3.3, TensorFlow 2.2.3 and TensorFlow 2.1.4, as these are also affected and still in supported range.