Papers by Chuan Sheng Foo
Using Punctuation as an Adversarial Attack on Deep Learning-Based NLP Systems: An Empirical Study (2023.findings-eacl)
Copied to clipboard
| Challenge: | Existing studies show that insertions of a few symbols are a general attack mechanism, but grammar checks do not mitigate them. |
| Approach: | They propose to use punctuation insertions as adversarial attacks on NLP systems to create a toolbox of methods to attack models while also pointing out flaws. |
| Outcome: | The results show that punctuation insertions outperform word-level attacks in settings with a limited number of word synonyms and queries to the victim’s model. |