PYSEC-2026-993

Impact

The implementation of tf.raw_ops.GetSessionTensor does not fully validate the input arguments. This results in a CHECK-failure which can be used to trigger a denial of service attack:

import tensorflow as tf

handle = tf.constant("[]", shape=[0], dtype=tf.string)
tf.raw_ops.GetSessionTensor(handle=handle)

The code assumes handle is a scalar but there is no validation for this:

    const Tensor& handle = ctx->input(0);
    const string& name = handle.scalar<tstring>()();

Patches

We have patched the issue in GitHub commit 48305e8ffe5246d67570b64096a96f8e315a7281.

The fix will be included in TensorFlow 2.9.0. We will also cherrypick this commit on TensorFlow 2.8.1, TensorFlow 2.7.2, and TensorFlow 2.6.4, as these are also affected and still in supported range.

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 Neophytos Christou from Secure Systems Lab at Brown University.