FIX: Exploit trust_remote_code in prompt tuning #2896
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Resolves huggingface#2888 This is a test for a hypothetical exploit that would enable trust_remote_code (and thus RCE) when a user loads a malicious prompt tuning model. This is because PEFT would just pass the on the tokenizer_kwargs defined in the prompt tuning config unsanitzed, which means that if the tokenizer is also malicious, the malicious code would be executed. For this exploit to work, a user cannot load a model using PeftModel.from_pretrained as normal, because the tokenizer is only loaded in training mode. Although the attacker could set inference_mode=True in the adapter_config.json, that would still not work because prompt tuning methods cannot be loaded in inference mode. Therefore, the only way for the exploit to work would be if the user manually loads the model.
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
- moved new test to other file to group similar test - fix old test that was failing due to change
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Resolves #2888
This is a test for a hypothetical exploit that would enable
trust_remote_code(and thus RCE) when a user loads a malicious prompt tuning model. This is because PEFT would just pass the on thetokenizer_kwargsdefined in the prompt tuning config unsanitzed, which means that if the tokenizer is also malicious, the malicious code would be executed.For this exploit to work, a user cannot load a model using
PeftModel.from_pretrainedas normal, because the tokenizer is only loaded in training mode. Although the attacker could setinference_mode=Truein theadapter_config.json, that would still not work because prompt tuning methods cannot be loaded in inference mode. Therefore, the only way for the exploit to work would be if the user manually loads the model.