If you are using TF on Windows or a Linux distribution that is known to have this denial of service issue, you are strongly advised to upgrade to a supported distribution. Alternatively, you can disable TF’s support for `tf.sparse.cross` at runtime by setting `tf_sparsetools_enabled=False`. This will limit the use of TF with `tf.sparse.cross` to cases where the native C library implementation is not vulnerable. We have received a few reports of this issue triggering on Linux distributions. If you are using Linux and are also experiencing this denial of service, it can be mitigated by disabling the use of `tf.sparse.cross`. To do so, set `tf_sparsetools_enabled=False` in your `tf_config`. As a best practice, we recommend that you upgrade to a supported distribution whenever possible.
References
- https://github.com/tensorflow/tensorflow/issues/35997
- https://github.com/tensorflow/tensorflow/pull/3436
- https://github.com/tensorflow/tensorflow/issues/30814
***NOTICE***: If you are using TF on Windows or a Linux distribution that is known to have this denial of service issue, you are strongly advised to upgrade to a supported distribution. Alternatively, you can disable TF’s support for `tf.sparse.cross` at runtime by setting `tf_sparsetools_enabled=False`. This will limit the use of TF with `tf.sparse.cross` to cases where the native C library implementation is not vulnerable. We have received a few reports of this issue triggering on Linux distributions.
Timeline
Published on: 09/16/2022 23:15:00 UTC
Last modified on: 09/20/2022 14:49:00 UTC