VYPR
Get started
← All advisories
Moderate severityNVD Advisory· Published Nov 18, 2022· Updated Apr 22, 2025

Segfault in `CompositeTensorVariantToComponents` in Tensorflow

CVE-2022-41909

Description

TensorFlow is an open source platform for machine learning. An input encoded that is not a valid CompositeTensorVariant tensor will trigger a segfault in tf.raw_ops.CompositeTensorVariantToComponents. We have patched the issue in GitHub commits bf594d08d377dc6a3354d9fdb494b32d45f91971 and 660ce5a89eb6766834bdc303d2ab3902aef99d3d. The fix will be included in TensorFlow 2.11. We will also cherrypick this commit on TensorFlow 2.10.1, 2.9.3, and TensorFlow 2.8.4, as these are also affected and still in supported range.

Affected packages

Versions sourced from the GitHub Security Advisory.

PackageAffected versionsPatched versions
tensorflowPyPI
< 2.8.42.8.4
tensorflowPyPI
>= 2.9.0, < 2.9.32.9.3
tensorflowPyPI
>= 2.10.0, < 2.10.12.10.1
tensorflow-cpuPyPI
< 2.8.42.8.4
tensorflow-gpuPyPI
< 2.8.42.8.4
tensorflow-cpuPyPI
>= 2.9.0, < 2.9.32.9.3
tensorflow-gpuPyPI
>= 2.9.0, < 2.9.32.9.3
tensorflow-cpuPyPI
>= 2.10.0, < 2.10.12.10.1
tensorflow-gpuPyPI
>= 2.10.0, < 2.10.12.10.1

Affected products

1

Patches

2
bf594d08d377

[Security] Raise an exception when input to CompositeTensorVariantToComponents is not a valid CompositeTensorVariant tensor.

https://github.com/tensorflow/tensorflowJun XuSep 15, 2022via ghsa
commit
2 files changed · +17 −0
  • tensorflow/core/kernels/composite_tensor_ops.cc+4 −0 modified
    @@ -73,6 +73,10 @@ class CompositeTensorVariantToComponents : public OpKernel {
                                 "tensor, but got ",
                                 encoded_t.DebugString()));
     auto* encoded = encoded_t.flat<Variant>()(0).get<CompositeTensorVariant>();
+    OP_REQUIRES(context, encoded != nullptr,
+                errors::InvalidArgument("The input `encoded` is not a valid "
+                                        "CompositeTensorVariant tensor, got ",
+                                        encoded_t.DebugString()));
 
     // Check that the encoded TypeSpec is compatible with the expected TypeSpec.
     // For now, we just check that the class matches.
  • tensorflow/python/kernel_tests/composite_tensor_ops_test.py+13 −0 modified
    @@ -25,6 +25,7 @@
 from tensorflow.python.framework import test_util
 from tensorflow.python.ops import composite_tensor_ops
 from tensorflow.python.ops import gen_composite_tensor_ops
+from tensorflow.python.ops import gen_list_ops
 from tensorflow.python.ops import gradients_impl
 from tensorflow.python.ops import math_ops
 from tensorflow.python.ops import parsing_ops
@@ -97,6 +98,18 @@ def testDecodingEmptyNonScalarTensorError(self):
           metadata='',
           Tcomponents=[dtypes.int32])
 
+  def testDecodingInvalidEncodedInputError(self):
+    with self.assertRaisesRegex(errors.InvalidArgumentError,
+                                'not a valid CompositeTensorVariant tensor'):
+      self.evaluate(
+          gen_composite_tensor_ops.CompositeTensorVariantToComponents(
+              encoded=gen_list_ops.EmptyTensorList(
+                  element_dtype=dtypes.int32,
+                  element_shape=[1, 2],
+                  max_num_elements=2),
+              metadata='',
+              Tcomponents=[dtypes.int32]))
+
   def testRoundTripThroughTensorProto(self):
     value = ragged_factory_ops.constant([[1, 2], [3], [4, 5, 6]])
     encoded = composite_tensor_ops.composite_tensor_to_variants(value)
660ce5a89eb6

[Security] Add a check for empty variant tensor input to CompositeTensorVariantToComponents.

https://github.com/tensorflow/tensorflowJun XuSep 14, 2022via ghsa
commit
2 files changed · +20 −0
  • tensorflow/core/kernels/composite_tensor_ops.cc+6 −0 modified
    @@ -15,6 +15,7 @@ limitations under the License.
 
 #include "tensorflow/core/framework/op.h"
 #include "tensorflow/core/framework/op_kernel.h"
+#include "tensorflow/core/framework/op_requires.h"
 #include "tensorflow/core/framework/variant.h"
 #include "tensorflow/core/framework/variant_encode_decode.h"
 #include "tensorflow/core/kernels/composite_tensor_variant.h"
@@ -66,6 +67,11 @@ class CompositeTensorVariantToComponents : public OpKernel {
 
   void Compute(OpKernelContext* context) override {
     Tensor encoded_t = context->input(0);
+    OP_REQUIRES(
+        context, encoded_t.flat<Variant>().size() > 0,
+        errors::InvalidArgument("Input `encoded` must not be an empty variant "
+                                "tensor, but got ",
+                                encoded_t.DebugString()));
     auto* encoded = encoded_t.flat<Variant>()(0).get<CompositeTensorVariant>();
 
     // Check that the encoded TypeSpec is compatible with the expected TypeSpec.
  • tensorflow/python/kernel_tests/composite_tensor_ops_test.py+14 −0 modified
    @@ -18,11 +18,13 @@
 
 from tensorflow.python.eager import backprop
 from tensorflow.python.eager import context
+from tensorflow.python.framework import constant_op
 from tensorflow.python.framework import dtypes
 from tensorflow.python.framework import errors
 from tensorflow.python.framework import sparse_tensor
 from tensorflow.python.framework import test_util
 from tensorflow.python.ops import composite_tensor_ops
+from tensorflow.python.ops import gen_composite_tensor_ops
 from tensorflow.python.ops import gradients_impl
 from tensorflow.python.ops import math_ops
 from tensorflow.python.ops import parsing_ops
@@ -83,6 +85,18 @@ def testEncodingErrors(self, value, spec, message):
     with self.assertRaisesRegex(ValueError, message):
       composite_tensor_ops.composite_tensor_to_variants(value(), spec)
 
+  def testDecodingEmptyNonScalarTensorError(self):
+    if not context.executing_eagerly():
+      # Creating a variant tensor of an empty list is not allowed in eager mode.
+      return
+
+    with self.assertRaisesRegex(errors.InvalidArgumentError,
+                                'must not be an empty variant tensor'):
+      gen_composite_tensor_ops.CompositeTensorVariantToComponents(
+          encoded=constant_op.constant([], dtype=dtypes.variant),
+          metadata='',
+          Tcomponents=[dtypes.int32])
+
   def testRoundTripThroughTensorProto(self):
     value = ragged_factory_ops.constant([[1, 2], [3], [4, 5, 6]])
     encoded = composite_tensor_ops.composite_tensor_to_variants(value)

Vulnerability mechanics

Generated by null/stub on May 9, 2026. Inputs: CWE entries + fix-commit diffs from this CVE's patches. Citations validated against bundle.

References

6

News mentions

0

No linked articles in our index yet.