Use after free / memory leak in `CollectiveReduceV2`

Impact The async implementation of CollectiveReduceV2 suffers from a memory leak and a use after free: python import tensorflow as tf tf.rawops.CollectiveReduceV2 input=, groupsize=-10, -10, -10, groupkey=-10, -10, instancekey=-10, orderingtoken=, mergeop='Mul', finalop='Div' This occurs due to t...

Access to invalid memory during shape inference in `Cudnn*` ops

Impact The shape inference code for the Cudnn operations in TensorFlow can be tricked into accessing invalid memory, via a heap buffer overflow: python import tensorflow as tf @tf.function def func: return tf.rawops.CudnnRNNV3 input=0.1, 0.1, inputh=0.5, inputc=0.1, 0.1, 0.1, params=0.5, 0.5,...

GHSA-CQV6-3PHM-HCWX Access to invalid memory during shape inference in `Cudnn*` ops

Impact The shape inference code for the Cudnn operations in TensorFlow can be tricked into accessing invalid memory, via a heap buffer overflow: python import tensorflow as tf @tf.function def func: return tf.rawops.CudnnRNNV3 input=0.1, 0.1, inputh=0.5, inputc=0.1, 0.1, 0.1, params=0.5, 0.5,...

Segfault due to negative splits in `SplitV`

Impact The implementation of SplitV can trigger a segfault is an attacker supplies negative arguments: python import tensorflow as tf tf.rawops.SplitV value=tf.constant, sizesplits=-1, -2 ,axis=0, numsplit=2 This occurs whenever sizesplits contains more than one value and at least one value is...

GHSA-CPF4-WX82-GXP6 Segfault due to negative splits in `SplitV`

Impact The implementation of SplitV can trigger a segfault is an attacker supplies negative arguments: python import tensorflow as tf tf.rawops.SplitV value=tf.constant, sizesplits=-1, -2 ,axis=0, numsplit=2 This occurs whenever sizesplits contains more than one value and at least one value is...

GHSA-F54P-F6JP-4RHR Heap OOB in `FusedBatchNorm` kernels

Impact The implementation of FusedBatchNorm kernels is vulnerable to a heap OOB: python import tensorflow as tf tf.rawops.FusedBatchNormGrad ybackprop=tf.constanti for i in range9,shape=1,1,3,3,dtype=tf.float32 x=tf.constanti for i in range2,shape=1,1,1,2,dtype=tf.float32 scale=1,1,...

Heap OOB in `FusedBatchNorm` kernels

Impact The implementation of FusedBatchNorm kernels is vulnerable to a heap OOB: python import tensorflow as tf tf.rawops.FusedBatchNormGrad ybackprop=tf.constanti for i in range9,shape=1,1,3,3,dtype=tf.float32 x=tf.constanti for i in range2,shape=1,1,1,2,dtype=tf.float32 scale=1,1,...

GHSA-RG3M-HQC5-344V `SparseFillEmptyRows` heap OOB

Impact The implementation of SparseFillEmptyRows can be made to trigger a heap OOB access: python import tensorflow as tf data=tf.rawops.SparseFillEmptyRows indices=0,0,0,0,0,0, values='sssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss', denseshape=5,3, defaultvalue='o' This occurs...

`SparseFillEmptyRows` heap OOB

Impact The implementation of SparseFillEmptyRows can be made to trigger a heap OOB access: python import tensorflow as tf data=tf.rawops.SparseFillEmptyRows indices=0,0,0,0,0,0, values='sssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss', denseshape=5,3, defaultvalue='o' This occurs...

A use of uninitialized value vulnerability in Tensorflow

Impact TensorFlow's Grappler optimizer has a use of unitialized variable: cc const NodeDef dequeuenode; for const auto& trainnode : trainnodes if IsDequeueOptrainnode dequeuenode = trainnode; break; if dequeuenode ... If the trainnodes vector obtained from the saved model that gets optimized does...

GHSA-7R94-XV9V-63JW A use of uninitialized value vulnerability in Tensorflow

Impact TensorFlow's Grappler optimizer has a use of unitialized variable: cc const NodeDef dequeuenode; for const auto& trainnode : trainnodes if IsDequeueOptrainnode dequeuenode = trainnode; break; if dequeuenode ... If the trainnodes vector obtained from the saved model that gets optimized does...

Heap OOB in `SparseBinCount`

Impact The implementation of SparseBinCount is vulnerable to a heap OOB: python import tensorflow as tf tf.rawops.SparseBincount indices=0,1,2 values=0,-10000000 denseshape=1,1 size=1 weights=3,2,1 binaryoutput=False This is because of missing validation between the elements of the values argumen...

GHSA-374M-JM66-3VJ8 Heap OOB in `SparseBinCount`

Impact The implementation of SparseBinCount is vulnerable to a heap OOB: python import tensorflow as tf tf.rawops.SparseBincount indices=0,1,2 values=0,-10000000 denseshape=1,1 size=1 weights=3,2,1 binaryoutput=False This is because of missing validation between the elements of the values argumen...

GHSA-J8C8-67VP-6MX7 Arbitrary memory read in `ImmutableConst`

Impact The ImmutableConst operation in TensorFlow can be tricked into reading arbitrary memory contents: python import tensorflow as tf with open'/tmp/test','wb' as f: f.writeb'\xe2'128 data = tf.rawops.ImmutableConstdtype=tf.string,shape=3,memoryregionname='/tmp/test' printdata This is because t...

Arbitrary memory read in `ImmutableConst`

Impact The ImmutableConst operation in TensorFlow can be tricked into reading arbitrary memory contents: python import tensorflow as tf with open'/tmp/test','wb' as f: f.writeb'\xe2'128 data = tf.rawops.ImmutableConstdtype=tf.string,shape=3,memoryregionname='/tmp/test' printdata This is because t...

Code injection in `saved_model_cli`

Impact TensorFlow's savedmodelcli tool is vulnerable to a code injection as it calls eval on user supplied strings python def preprocessinputexprsargstringinputexprsstr: ... for inputraw in filterbool, inputexprsstr.split';': ... inputkey, expr = inputraw.split'=', 1 inputdictinputkey = evalexpr...

Denial Of Service (DoS)

tensorflow is vulnerable to denial of service DoS. The vulnerability exists because the API can be made to deadlock when two tf.function functions are mutually recursive by using Lock Python object, allowing an attacker to crash the application by calling a recursive tf.function...

Denial Of Service (DoS)

tensorflow is vulnerable to denial of service. The vulnerability exists because adding shape checks for GPU TridiagonalMatMul op is not properly validated allowing invalid writes and reads from heap populated arrays which results in an application crash...

Denial Of Service (DoS)

tensorflow is vulnerable to denial of service. The shape inference code for QuantizeV2 can trigger a read outside of bounds of the heap-allocated array when axis is a negative value less than -1, which allows an attacker to cause an application crash...

Denial Of Service (DoS)

tensorflow is vulnerable to denial of service. The shape inference code for DeserializeSparse can trigger a null pointer dereference, allowing an attacker to cause an application crash...