google tensorflow
431 known vulnerabilities across versions
Vulnerabilities are listed by affected version. Select any CVE for the full briefing and its intelligence graph.
array_ops.upper_bound causes a segfault when not given ann_ops.fractional_avg_pool_v2 and `tf.raw_ops.LookupTableImportV2 cannot handle scalars inSparseSparseMaximum is givctx-step_containter() is asummarize ofindices for `Dconst char* array (sinencoded that is not a valid CompositeTensorVariant tensortoken that is not a UTF-8 bytestring will trigger a CHECKtf.raw_ops.ResizeNearestNeighborGrad is given a large size insparse_matrix that is not a matrix with a shape with rank 0dense_features or example_state_data not of rank 2 will triSparseFillEmptyRowsGrad is given empty inputs, TensorFlow will crFractionMaxPoolGrad is given outsize inputs `row_pooling_sequenceThreadUnsafeUnigramCandidateSampler is given input `filterbank_chMirrorPadGrad is given outsize input paddings, TensorFlow willCONV_3D_TRANSPOSE TensorFlow Lite operattf.raw_ops.TensorListResize is given a nonscalar value for inputtf.raw_ops.TensorListConcat is given element_shape=[], it resulBCast::ToShape is given input larger than an int32, it will cratf.image.generate_bounding_box_proposals receivtf.keras.losses.poisson receives a y_pred and y_true that are patf.raw_ops.ImageProjectiveTransformV2 is given a large output stf.raw_ops.FusedResizeAndPadConv2D is given a large tensor shapBaseCandidateSamplerOp function receives a value in `true_cRequantize is given input_min, input_max, `requested_output_mtensorflow::full_type::SubstituteFromAttrs receives a `FullTypeRangeSize receives values that do not fit into an int64_t, itmlir::tfg::TFOp::nameAttr receives null type list attributes, imlir::tfg::GraphDefImporter::ConvertNodeDef tries to convert Nomlir::tfg::ConvertGenericFunctionToFunctionDef is given empty fmlir::tfg::ConvertGenericFunctionToFunctionDef is given empty ftf.quantization.fake_quant_with_min_max_vars_gradient receivestf.random.gamma receives large input shape and rates, it givesRandomPoissonV2 receives large input shape and rates, it givesUnbatch receives a nonscalar input id, it gives a CHECK faiDrawBoundingBoxes receives an input boxes that is not of dtypmlir::tfg::ConvertGenericFunctionToFunctionDef is given empty fConv2DBackpropInput receives empty out_backprop inputs (e.g.EmptyTensorList receives an input element_shape with more thantf.sparse.cross receives an input separator that is not a scalaConv2D is given empty input and the filter and padding sizeAudioSummaryV2 receives an input sample_rate with more than oCollectiveGather receives an scalar input input, it gives a `SetSize receives an input set_shape that is not a 1D tensor,TensorListFromTensor receives an element_shape of a rank greaTensorListScatter and TensorListScatterV2 receive an `elementQuantizeAndDequantizeV3 is given a nonscalar num_bits input tenFakeQuantWithMinMaxVarsPerChannel is given min or max tensorsRaggedTensorToVariant is given a rt_nested_splits list that conMaxPool receives a window size input array ksize with dimensitf.linalg.matrix_rank receives an empty input a, the GPU kernDenseBincount assumes its input tensor weights to either have theRaggedBincount is given an empty input tensor splits, it resultLRNGrad is given an output_image input tensor that is not 4-D,ParameterizedTruncatedNormal assumes shape is of type int32. A vSave or SaveSlices is run over tensors of an unsupported `dtypeSparseBincount is given inputs for indices, values, and `densFractionalMaxPoolGrad validates its inputs with CHECK failures insQuantizedRelu or QuantizedRelu6 are given nonscalar inputs forQuantizeDownAndShrinkRange is given nonscalar inputs for `input_mQuantizedMatMul is given nonscalar input for: min_a, max_a, `QuantizedBiasAdd is given min_input, max_input, min_bias, `FakeQuantWithMinMaxVars is given min or max tensors of a nonzQuantizedInstanceNorm is given x_min or x_max tensors of a noConv2DBackpropInput requires input_sizes toAvgPoolGrad does not fully validate the inputQuantizedAdd is given min_input or max_input tensors of a nonQuantizedAvgPool is given min_input or max_input tensors of aLowerBound or UpperBound is given an emptysorted_inputs inputBlockLSTMGradV2 does not fully validate its inFractionalAvgPoolGrad does not fully validatecore/kernels/list_kernels.cc's TensorListReserve, num_elementsAvgPool3DGradOp does not fully validate the inUnbatchGradOp function takes an argument id that is assumed toAvgPoolOp function takes an argument ksize that must be positiRaggedRangOp function takes an argument limits that is eventuaScatterNd function takes an input argument that determines the iGatherNd function takes arguments that determine the sizes of inGatherNd function takes arguments that determine the sizes of inTensorKey hash function used total estimated `simplifyBroadcast function in the MLIR-TFRT infrastructure in TensoGraphDef format in TensorFlow does not allow self recursive functioGraphDef from a TensorFlow SavedModel can be maliciously altered toSavedModelSavedModelSavedModelSavedModelGetInitOp is vulnerable to a crash caused by dereOpLevelCostEstimator::CalculateOutputSize is vulnOpLevelCostEstimator::CalculateTensorSize is vulnSpecializeType which results in heap OOAssignOp can result in copying uninitialized dataGrappler. The `set_outptempfile.mktemp to create temporaryRange suffers from integer overflows. These can tSparseCountSparseOutput is vulnerable to a heap oQuantizedMaxPool has an undefined behavior whereSparseCountSparseOutput can be made to crash a Te*Bincount operations allows malicious users to caFractionalMaxPool can be made to crash a TensorFlMapStage is vulnerable a CHECK-fail if the keyUnravelIndex is vulnerable to a division by zeroAddManySparseToTensorsMap is vulnerable to an intSparseCwise ops are vulnerable to integer overSparseTensorSliceDataset has an undefined behavioStringNGrams can be used to trigger a denial of sThreadPoolHandle can be used to trigger a denialConcatV2 can be used to triggFractionalAvgPoolGrad does not consider cases wheReverseSequence does not fullDequantize is vulnerable to aDequantize does not fully validate the value of `saved_model_cli tool is vulnerableImmutableConst operation in TensorFlow canSplitV can trigger a segfCudnn* operatiCollectiveReduceV2Transpose is vtf.function API can be made toAllToAll can be maParallelConcat misses somSparseBinCount is vulneraSparseFillEmptyRows can bFusedBatchNorm kernels isDeserializeSparsetf.ragged.cross hatf.ragged.cross caQuantizeV2 can triEinsumHelper::ParseEquation()tf.summary.create_file_writer is called withtf.image.resize is called with a large inputtf.tile is called with a large input argumentf.math.segment_* operatiGatherNd implementation doeexpand_dims.cc contains a vtf.map_fn wtf.raw_ops.SaveV2 doestf.raw_ops.UncompressElement can be made torow_partition_types of `tf.rtf.transpose at the same tiCHECK-failParseAttrValue(https://github.com/ttf.io.decode_raw produces incorrecttf.raw_ops.CTCLoss allows an attSparseReshape results in a deniatf.raw_ops.QuantizeAndDequantizeV2 alloSparseAdd results in allowing atSparseAdd results in allowing atTFLiteIntArrays is vulneraDepthwiseConv TFLite operator iOneHot TFLite operator is vulneSplit TFLite operator is vulnerSVDF TFLite operator is vulneraSpaceToBatchNd TFLite operatorEmbeddingLookup TFLite operatorDepthToSpace TFLite operator isBatchToSpaceNd TFLite operatorMinimum and Maximum TFLite oGatherNd TFLite operaTransposeConv TFLitePrepare step of the SpaceToDepth TFLite operator doCHECK-ftf.raw_ops.FusedBatchNorm is vulnertf.raw_ops.Dequantize, an attf.raw_ops.FractionalMaxPoolGrad trtf.raw_ops.MaxPoolGrad is vulnerabltf.raw_ops.FractionalAvgPoolGrad istf.raw_ops.AvgPool3DGrad is vulneratf.raw_ops.MaxPool3DGradGrad is vultf.raw_ops.ReverseSequence allows ftf.raw_ops.MaxPool3DGradGrad exhibitf.raw_ops.MaxPoolGradWithArgmax istf.raw_ops.SdcaOptimizer triggers utf.raw_ops.MaxPoolGradWithArgmax catf.raw_ops.MaxPoolGradWithArgmax catf.raw_ops.MaxPoolGradWithArgmax caMatrixTriangularSolve(https://githuCHECK-fCHECK-fCHECK-ftf.raw_ops.ImmutableConst(https://www.tensorflow.CHECK-fCHECK fCHECK fail in PNG encoding by pCHECK-ftf.raw_ops.Conv3DBackprop* operations fail to validattf.raw_ops.SparseCross allows combinations whiConv3Dtf.raw_ops.RaggedTensorToVariant with arguments sMatrixDiag* operations(https://githsplits argument of RaggedBincount does not specisplits argument of RaggedBincount does not speciboxes argument of tf.image.crop_and_resize has a very large value, the CPU kernelaxis value to tf.quantization.quantize_and_dequantize. ThiSavedModel protocol buffer and aldata_splits argument of tf.raw_ops.StringNGrams lacksfill argument of tf.strings.as_string, aShard API in TensorFlow expects the last argument to beRaggedCountSparseOutput implementation does not validate that the input arguments form aRaggedCountSparseOutput implementation does not validate that the input arguments form aRaggedCountSparseOutput does not validate that the input arguments form a valid ragged tSparseCountSparseOutput implementation does not validate that the input arguments form aSparseCountSparseOutput implementation does not validate that the input arguments form aSparseCountSparseOutput and RaggedCountSparseOutput implementations don't validate that the `SparseFillEmptyRowsGrad uses a doubleSparseFillEmptyRowsGrad implementation has incomplete vdlpack.to_dlpack can be made to use uninitialized memory redlpack.to_dlpack there is a memory leak foldlpack.to_dlpack the expected validationstf.raw_ops.Switch operation takes as input a tensor and