TensorFlow is an end-to-end open source platform for machine learning. An attacker can cause a denial of service via a FPE runtime error in `tf.raw_ops.DenseCountSparseOutput`. This is because the implementation(https://github.com/tensorflow/tensorflow/blob/efff014f3b2d8ef6141da30c806faf141297eca1/tensorflow/core/kernels/count_ops.cc#L123-L127) computes a divisor value from user data but does not check that the result is 0 before doing the division. Since `data` is given by the `values` argument, `num_batch_elements` is 0. The fix will be included in TensorFlow 2.5.0. We will also cherrypick this commit on TensorFlow 2.4.2, and TensorFlow 2.3.3, as these are also affected.

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.

In affected versions of TensorFlow the tf.raw_ops.ImmutableConst operation returns a constant tensor created from a memory mapped file which is assumed immutable. However, if the type of the tensor is not an integral type, the operation crashes the Python interpreter as it tries to write to the memory area. If the file is too small, TensorFlow properly returns an error as the memory area has fewer bytes than what is needed for the tensor it creates. However, as soon as there are enough bytes, the above snippet causes a segmentation fault. This is because the allocator used to return the buffer data is not marked as returning an opaque handle since the needed virtual method is not overridden. This is fixed in versions 1.15.5, 2.0.4, 2.1.3, 2.2.2, 2.3.2, and 2.4.0.

In affected versions of TensorFlow the tf.raw_ops.DataFormatVecPermute API does not validate the src_format and dst_format attributes. The code assumes that these two arguments define a permutation of NHWC. This can result in uninitialized memory accesses, read outside of bounds and even crashes. This is fixed in versions 1.15.5, 2.0.4, 2.1.3, 2.2.2, 2.3.2, and 2.4.0.

In affected versions of TensorFlow under certain cases a saved model can trigger use of uninitialized values during code execution. This is caused by having tensor buffers be filled with the default value of the type but forgetting to default initialize the quantized floating point types in Eigen. This is fixed in versions 1.15.5, 2.0.4, 2.1.3, 2.2.2, 2.3.2, and 2.4.0.

In affected versions of TensorFlow under certain cases, loading a saved model can result in accessing uninitialized memory while building the computation graph. The MakeEdge function creates an edge between one output tensor of the src node (given by output_index) and the input slot of the dst node (given by input_index). This is only possible if the types of the tensors on both sides coincide, so the function begins by obtaining the corresponding DataType values and comparing these for equality. However, there is no check that the indices point to inside of the arrays they index into. Thus, this can result in accessing data out of bounds of the corresponding heap allocated arrays. In most scenarios, this can manifest as unitialized data access, but if the index points far away from the boundaries of the arrays this can be used to leak addresses from the library. This is fixed in versions 1.15.5, 2.0.4, 2.1.3, 2.2.2, 2.3.2, and 2.4.0.

In Tensorflow before version 2.4.0, when the `boxes` argument of `tf.image.crop_and_resize` has a very large value, the CPU kernel implementation receives it as a C++ `nan` floating point value. Attempting to operate on this is undefined behavior which later produces a segmentation fault. The issue is patched in eccb7ec454e6617738554a255d77f08e60ee0808 and TensorFlow 2.4.0 will be released containing the patch. TensorFlow nightly packages after this commit will also have the issue resolved.

In Tensorflow before version 2.4.0, an attacker can pass an invalid `axis` value to `tf.quantization.quantize_and_dequantize`. This results in accessing a dimension outside the rank of the input tensor in the C++ kernel implementation. However, dim_size only does a DCHECK to validate the argument and then uses it to access the corresponding element of an array. Since in normal builds, `DCHECK`-like macros are no-ops, this results in segfault and access out of bounds of the array. The issue is patched in eccb7ec454e6617738554a255d77f08e60ee0808 and TensorFlow 2.4.0 will be released containing the patch. TensorFlow nightly packages after this commit will also have the issue resolved.

In Tensorflow before versions 1.15.4, 2.0.3, 2.1.2, 2.2.1 and 2.3.1, changing the TensorFlow's `SavedModel` protocol buffer and altering the name of required keys results in segfaults and data corruption while loading the model. This can cause a denial of service in products using `tensorflow-serving` or other inference-as-a-service installments. Fixed were added in commits f760f88b4267d981e13f4b302c437ae800445968 and fcfef195637c6e365577829c4d67681695956e7d (both going into TensorFlow 2.2.0 and 2.3.0 but not yet backported to earlier versions). However, this was not enough, as #41097 reports a different failure mode. The issue is patched in commit adf095206f25471e864a8e63a0f1caef53a0e3a6, and is released in TensorFlow versions 1.15.4, 2.0.3, 2.1.2, 2.2.1, or 2.3.1.

In Tensorflow before versions 1.15.4, 2.0.3, 2.1.2, 2.2.1 and 2.3.1, the `data_splits` argument of `tf.raw_ops.StringNGrams` lacks validation. This allows a user to pass values that can cause heap overflow errors and even leak contents of memory In the linked code snippet, all the binary strings after `ee ff` are contents from the memory stack. Since these can contain return addresses, this data leak can be used to defeat ASLR. The issue is patched in commit 0462de5b544ed4731aa2fb23946ac22c01856b80, and is released in TensorFlow versions 1.15.4, 2.0.3, 2.1.2, 2.2.1, or 2.3.1.

In eager mode, TensorFlow before versions 1.15.4, 2.0.3, 2.1.2, 2.2.1 and 2.3.1 does not set the session state. Hence, calling `tf.raw_ops.GetSessionHandle` or `tf.raw_ops.GetSessionHandleV2` results in a null pointer dereference In linked snippet, in eager mode, `ctx->session_state()` returns `nullptr`. Since code immediately dereferences this, we get a segmentation fault. The issue is patched in commit 9a133d73ae4b4664d22bd1aa6d654fec13c52ee1, and is released in TensorFlow versions 1.15.4, 2.0.3, 2.1.2, 2.2.1, or 2.3.1.

In Tensorflow before versions 1.15.4, 2.0.3, 2.1.2, 2.2.1 and 2.3.1, by controlling the `fill` argument of tf.strings.as_string, a malicious attacker is able to trigger a format string vulnerability due to the way the internal format use in a `printf` call is constructed. This may result in segmentation fault. The issue is patched in commit 33be22c65d86256e6826666662e40dbdfe70ee83, and is released in TensorFlow versions 1.15.4, 2.0.3, 2.1.2, 2.2.1, or 2.3.1.

In Tensorflow before versions 1.15.4, 2.0.3, 2.1.2, 2.2.1 and 2.3.1, the `Shard` API in TensorFlow expects the last argument to be a function taking two `int64` (i.e., `long long`) arguments. However, there are several places in TensorFlow where a lambda taking `int` or `int32` arguments is being used. In these cases, if the amount of work to be parallelized is large enough, integer truncation occurs. Depending on how the two arguments of the lambda are used, this can result in segfaults, read/write outside of heap allocated arrays, stack overflows, or data corruption. The issue is patched in commits 27b417360cbd671ef55915e4bb6bb06af8b8a832 and ca8c013b5e97b1373b3bb1c97ea655e69f31a575, and is released in TensorFlow versions 1.15.4, 2.0.3, 2.1.2, 2.2.1, or 2.3.1.

In Tensorflow before versions 1.15.4, 2.0.3, 2.1.2, 2.2.1 and 2.3.1, the implementation of `SparseFillEmptyRowsGrad` uses a double indexing pattern. It is possible for `reverse_index_map(i)` to be an index outside of bounds of `grad_values`, thus resulting in a heap buffer overflow. The issue is patched in commit 390611e0d45c5793c7066110af37c8514e6a6c54, and is released in TensorFlow versions 1.15.4, 2.0.3, 2.1.2, 2.2.1, or 2.3.1.

In Tensorflow before versions 1.15.4, 2.0.3, 2.1.2, 2.2.1 and 2.3.1, the `SparseFillEmptyRowsGrad` implementation has incomplete validation of the shapes of its arguments. Although `reverse_index_map_t` and `grad_values_t` are accessed in a similar pattern, only `reverse_index_map_t` is validated to be of proper shape. Hence, malicious users can pass a bad `grad_values_t` to trigger an assertion failure in `vec`, causing denial of service in serving installations. The issue is patched in commit 390611e0d45c5793c7066110af37c8514e6a6c54, and is released in TensorFlow versions 1.15.4, 2.0.3, 2.1.2, 2.2.1, or 2.3.1."

In Tensorflow before versions 1.15.4, 2.0.3, 2.1.2, 2.2.1 and 2.3.1, the `tf.raw_ops.Switch` operation takes as input a tensor and a boolean and outputs two tensors. Depending on the boolean value, one of the tensors is exactly the input tensor whereas the other one should be an empty tensor. However, the eager runtime traverses all tensors in the output. Since only one of the tensors is defined, the other one is `nullptr`, hence we are binding a reference to `nullptr`. This is undefined behavior and reported as an error if compiling with `-fsanitize=null`. In this case, this results in a segmentation fault The issue is patched in commit da8558533d925694483d2c136a9220d6d49d843c, and is released in TensorFlow versions 1.15.4, 2.0.3, 2.1.2, 2.2.1, or 2.3.1.

In TensorFlow before 1.15.2 and 2.0.1, converting a string (from Python) to a tf.float16 value results in a segmentation fault in eager mode as the format checks for this use case are only in the graph mode. This issue can lead to denial of service in inference/training where a malicious attacker can send a data point which contains a string instead of a tf.float16 value. Similar effects can be obtained by manipulating saved models and checkpoints whereby replacing a scalar tf.float16 value with a scalar string will trigger this issue due to automatic conversions. This can be easily reproduced by tf.constant("hello", tf.float16), if eager execution is enabled. This issue is patched in TensorFlow 1.15.1 and 2.0.1 with this vulnerability patched. TensorFlow 2.1.0 was released after we fixed the issue, thus it is not affected. Users are encouraged to switch to TensorFlow 1.15.1, 2.0.1 or 2.1.0.

In TensorFlow before 1.15, a heap buffer overflow in UnsortedSegmentSum can be produced when the Index template argument is int32. In this case data_size and num_segments fields are truncated from int64 to int32 and can produce negative numbers, resulting in accessing out of bounds heap memory. This is unlikely to be exploitable and was detected and fixed internally in TensorFlow 1.15 and 2.0.

NULL pointer dereference in Google TensorFlow before 1.12.2 could cause a denial of service via an invalid GIF file.