Impact

Changing the TensorFlow's SavedModel protocol buffer and altering the name of required keys results in segfaults and data corruption while loading the model. This can cause a denial of service in products using tensorflow-serving or other inference-as-a-service installments.

We have added fixes to this in f760f88 and fcfef19 (both going into TensorFlow 2.2.0 and 2.3.0 but not yet backported to earlier versions). However, this was not enough, as #41097 reports a different failure mode.

Patches

We have patched the issue in adf0952 and will release patch releases for all versions between 1.15 and 2.3. Patch releases for versions between 1.15 and 2.1 will also contain cherry-picks of f760f88 and fcfef19.

We recommend users to upgrade to TensorFlow 1.15.4, 2.0.3, 2.1.2, 2.2.1, or 2.3.1.

For more information

Please consult our security guide for more information regarding the security model and how to contact us with issues and questions.

Attribution

This vulnerability has been reported by Shuaike Dong, from Alipay Tian Qian Security Lab && Lab for Applied Security Research, CUHK.