tfsa-2023-008.md

TFSA-2023-008: FPE in AudioSpectrogram

CVE Number

CVE-2023-25666

Impact

version:2.11.0 audio_ops.cc:70

Status SpectrogramShapeFn(InferenceContext* c) { ShapeHandle input; TF_RETURN_IF_ERROR(c->WithRank(c->input(0), 2, &input)); int32_t window_size; TF_RETURN_IF_ERROR(c->GetAttr("window_size", &window_size)); int32_t stride; TF_RETURN_IF_ERROR(c->GetAttr("stride", &stride)); .....[1]

DimensionHandle input_length = c->Dim(input, 0); DimensionHandle input_channels = c->Dim(input, 1);

DimensionHandle output_length; if (!c->ValueKnown(input_length)) { output_length = c->UnknownDim(); } else { const int64_t input_length_value = c->Value(input_length); const int64_t length_minus_window = (input_length_value - window_size); int64_t output_length_value; if (length_minus_window < 0) { output_length_value = 0; } else { output_length_value = 1 + (length_minus_window / stride); .....[2] } output_length = c->MakeDim(output_length_value); }

Get the value of stride at [1], and the used at [2]

import tensorflow as tf

para = {'input': tf.constant([[14.], [24.]], dtype=tf.float32), 'window_size': 1, 'stride': 0, 'magnitude_squared': False}
func = tf.raw_ops.AudioSpectrogram

@tf.function(jit_compile=True)
def fuzz_jit():
   y = func(**para)
   return y

fuzz_jit()

Patches

We have patched the issue in GitHub commit d0d4e779da0d0f56499c6fa5ba09f0a576cc6b14.

The fix will be included in TensorFlow 2.12.0. We will also cherrypick this commit on TensorFlow 2.11.1

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 r3pwnx