Code injection in `saved_model_cli` in TensorFlow

ImpactTensorFlow's savedmodelcli tool is vulnerable to a code injection:savedmodelcli run --inputexprs 'x=print"malicious code to run"' --dir ./--tagset serve --signaturedef servingdefaultThis can be used to open a reverse shell savedmodelcli run --inputexprs 'hello=exec"""\nimport...

TensorFlow vulnerable to `CHECK` fail in `Conv2DBackpropInput`

ImpactWhen Conv2DBackpropInput receives empty outbackprop inputs e.g. 3, 1, 0, 1, the current CPU/GPU kernels CHECK fail one with dnnl, the other with cudnn. This can be used to trigger a denial of service attack.pythonimport tensorflow as tfimport numpy as npinputsizes = 3, 1, 1, 2filter =...

TensorFlow vulnerable to `CHECK` fail in `Conv2DBackpropInput`

ImpactWhen Conv2DBackpropInput receives empty outbackprop inputs e.g. 3, 1, 0, 1, the current CPU/GPU kernels CHECK fail one with dnnl, the other with cudnn. This can be used to trigger a denial of service attack.pythonimport tensorflow as tfimport numpy as npinputsizes = 3, 1, 1, 2filter =...

PYSEC-2026-3289 Code injection in `saved_model_cli` in TensorFlow

Impact TensorFlow's savedmodelcli tool is vulnerable to a code injection: savedmodelcli run --inputexprs 'x=print"malicious code to run"' --dir ./ --tagset serve --signaturedef servingdefault This can be used to open a reverse shell savedmodelcli run --inputexprs 'hello=exec"""\nimport...

PYSEC-2026-3269 TensorFlow vulnerable to `CHECK` fail in `Conv2DBackpropInput`

Impact When Conv2DBackpropInput receives empty outbackprop inputs e.g. 3, 1, 0, 1, the current CPU/GPU kernels CHECK fail one with dnnl, the other with cudnn. This can be used to trigger a denial of service attack. python import tensorflow as tf import numpy as np inputsizes = 3, 1, 1, 2 filter =...

Incomplete validation in signal ops leads to crashes in TensorFlow

ImpactThe tf.compat.v1.signal.rfft2d and tf.compat.v1.signal.rfft3d lack input validation and under certain condition can result in crashes due to CHECK-failures. Patches We have patched the issue in GitHub commit 0a8a781e597b18ead006d19b7d23d0a369e9ad73 merging GitHub PR 55274.The fix will be...

PYSEC-2026-3125 Code injection in `saved_model_cli` in TensorFlow

Impact TensorFlow's savedmodelcli tool is vulnerable to a code injection: savedmodelcli run --inputexprs 'x=print"malicious code to run"' --dir ./ --tagset serve --signaturedef servingdefault This can be used to open a reverse shell savedmodelcli run --inputexprs 'hello=exec"""\nimport...

PYSEC-2026-3299 Core dump when loading TFLite models with quantization in TensorFlow

Impact Certain TFLite models that were created using TFLite model converter would crash when loaded in the TFLite interpreter. The culprit is that during quantization the scale of values could be greater than 1 but code was always assuming sub-unit scaling. Thus, since code was calling...

PYSEC-2026-3141 Core dump when loading TFLite models with quantization in TensorFlow

Impact Certain TFLite models that were created using TFLite model converter would crash when loaded in the TFLite interpreter. The culprit is that during quantization the scale of values could be greater than 1 but code was always assuming sub-unit scaling. Thus, since code was calling...

PYSEC-2026-3107 Incomplete validation in signal ops leads to crashes in TensorFlow

Impact The tf.compat.v1.signal.rfft2d and tf.compat.v1.signal.rfft3d lack input validation and under certain condition can result in crashes due to CHECK-failures. Patches We have patched the issue in GitHub commit 0a8a781e597b18ead006d19b7d23d0a369e9ad73 merging GitHub PR 55274. The fix will be...

PYSEC-2026-3092 TensorFlow vulnerable to `CHECK` fail in `Conv2DBackpropInput`

Impact When Conv2DBackpropInput receives empty outbackprop inputs e.g. 3, 1, 0, 1, the current CPU/GPU kernels CHECK fail one with dnnl, the other with cudnn. This can be used to trigger a denial of service attack. python import tensorflow as tf import numpy as np inputsizes = 3, 1, 1, 2 filter =...

Type confusion leading to `CHECK`-failure based denial of service in TensorFlow

ImpactThe macros that TensorFlow uses for writing assertions e.g., CHECKLT, CHECKGT, etc. have an incorrect logic when comparing sizet and int values. Due to type conversion rules, several of the macros would trigger incorrectly. PatchesWe have patched the issue in GitHub commit...

Type confusion leading to `CHECK`-failure based denial of service in TensorFlow

ImpactThe macros that TensorFlow uses for writing assertions e.g., CHECKLT, CHECKGT, etc. have an incorrect logic when comparing sizet and int values. Due to type conversion rules, several of the macros would trigger incorrectly. PatchesWe have patched the issue in GitHub commit...

Heap buffer overflow due to incorrect hash function in TensorFlow

ImpactThe TensorKey hash function used total estimated AllocatedBytes, which a is an estimate per tensor, and b is a very poor hash function for constants e.g. int32t. It also tried to access individual tensor bytes through tensor.data of size AllocatedBytes. This led to ASAN failures because the...

Heap buffer overflow due to incorrect hash function in TensorFlow

ImpactThe TensorKey hash function used total estimated AllocatedBytes, which a is an estimate per tensor, and b is a very poor hash function for constants e.g. int32t. It also tried to access individual tensor bytes through tensor.data of size AllocatedBytes. This led to ASAN failures because the...

Missing validation results in undefined behavior in `SparseTensorDenseAdd

ImpactThe implementation of tf.rawops.SparseTensorDenseAdd does not fully validate the input arguments:pythonimport tensorflow as tfaindices = tf.constant0, shape=17, 2, dtype=tf.int64avalues = tf.constant, shape=0, dtype=tf.float32ashape = tf.constant6, 12, shape=2, dtype=tf.int64b =...

Segfault if `tf.histogram_fixed_width` is called with NaN values in TensorFlow

ImpactThe implementation of tf.histogramfixedwidth is vulnerable to a crash when the values array contain NaN elements:pythonimport tensorflow as tfimport numpy as nptf.histogramfixedwidthvalues=np.nan, valuerange=1,2The implementation assumes that all floating point operations are defined and th...

Segfault and OOB write due to incomplete validation in `EditDistance` in TensorFlow

ImpactThe implementation of tf.rawops.EditDistance has incomplete validation. Users can pass negative values to cause a segmentation fault based denial of service:pythonimport tensorflow as tfhypothesisindices = tf.constant-1250999896764, shape=3, 3, dtype=tf.int64 hypothesisvalues = tf.constant0...

Undefined behavior when users supply invalid resource handles

ImpactMultiple TensorFlow operations misbehave in eager mode when the resource handle provided to them is invalid:pythonimport tensorflow as tftf.rawops.QueueIsClosedV2handle=pythonimport tensorflow as tftf.summary.flushwriter= In graph mode, it would have been impossible to perform these API...

Undefined behavior when users supply invalid resource handles

ImpactMultiple TensorFlow operations misbehave in eager mode when the resource handle provided to them is invalid:pythonimport tensorflow as tftf.rawops.QueueIsClosedV2handle=pythonimport tensorflow as tftf.summary.flushwriter= In graph mode, it would have been impossible to perform these API...