Headline
CVE-2022-41885: Fix tensor shape overflow in FusedResizeAndPadConv2D. · tensorflow/tensorflow@d66e1d5
TensorFlow is an open source platform for machine learning. When tf.raw_ops.FusedResizeAndPadConv2D is given a large tensor shape, it overflows. We have patched the issue in GitHub commit d66e1d568275e6a2947de97dca7a102a211e01ce. The fix will be included in TensorFlow 2.11. We will also cherrypick this commit on TensorFlow 2.10.1, 2.9.3, and TensorFlow 2.8.4, as these are also affected and still in supported range.
@@ -3429,6 +3429,33 @@ def testAddWithSameSrcAndAddTensorBuffer(self):
np.rint(expected_output),
self.evaluate(add).reshape(-1))
# Fused resize and pad conv.
@test_util.run_in_graph_and_eager_modes()
def testResizeAndPadLargeResize(self):
with self.assertRaisesRegex((ValueError, errors_impl.InvalidArgumentError),
“Encountered overflow”):
mode = “REFLECT”
strides = [1, 1, 1, 1]
padding = “SAME”
resize_align_corners = False
tensor = constant_op.constant(
147, shape=[3, 3, 1, 4], dtype=dtypes.float32)
size = constant_op.constant([1879048192, 1879048192], dtype=dtypes.int32)
paddings = constant_op.constant([[0, 0], [0, 0], [0, 0], [0, 0]],
dtype=dtypes.int32)
kernel = constant_op.constant(
123, shape=[1, 3, 4, 1], dtype=dtypes.float32)
self.evaluate(
gen_nn_ops.fused_resize_and_pad_conv2d(
input=tensor,
size=size,
paddings=paddings,
filter=kernel,
mode=mode,
strides=strides,
padding=padding,
resize_align_corners=resize_align_corners))
if __name__ == "__main__":
for index, (input_size_, filter_size_, output_size_, stride_,
Related news
### Impact When [`tf.raw_ops.FusedResizeAndPadConv2D`](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/kernels/conv_ops_fused_image_transform.cc) is given a large tensor shape, it overflows. ```python import tensorflow as tf mode = "REFLECT" strides = [1, 1, 1, 1] padding = "SAME" resize_align_corners = False input = tf.constant(147, shape=[3,3,1,1], dtype=tf.float16) size = tf.constant([1879048192,1879048192], shape=[2], dtype=tf.int32) paddings = tf.constant([3,4], shape=[2], dtype=tf.int32) filter = tf.constant(123, shape=[1,3,4,1], dtype=tf.float16) tf.raw_ops.FusedResizeAndPadConv2D(input=input, size=size, paddings=paddings, filter=filter, mode=mode, strides=strides, padding=padding, resize_align_corners=resize_align_corners) ``` ### Patches We have patched the issue in GitHub commit [d66e1d568275e6a2947de97dca7a102a211e01ce](https://github.com/tensorflow/tensorflow/commit/d66e1d568275e6a2947de97dca7a102a211e01ce). The fix will be included in TensorFlow 2.1...