In affected versions of TensorFlow running an LSTM/GRU model where the LSTM/GRU layer receives an input with zero-length results in a CHECK failure when using the CUDA backend. This can result in a query-of-death vulnerability, via denial of service, if users can control the input to the layer. This is fixed in versions 1.15.5, 2.0.4, 2.1.3, 2.2.2, 2.3.2, and 2.4.0.
[
{
"signature_type": "Line",
"digest": {
"threshold": 0.9,
"line_hashes": [
"166252882886534993164129206683886691984",
"82771836372170310932193629693836048928",
"130140444966903475628418281477362812430",
"31228978155751701708215333570039799818",
"71908040881237060325670982965864455659",
"165435079343752882710408888814154468784",
"130140444966903475628418281477362812430",
"31228978155751701708215333570039799818"
]
},
"target": {
"file": "tensorflow/stream_executor/cuda/cuda_dnn.cc"
},
"signature_version": "v1",
"id": "CVE-2020-26270-36cb9890",
"deprecated": false,
"source": "https://github.com/tensorflow/tensorflow/commit/14755416e364f17fb1870882fa778c7fec7f16e3"
}
]