Certified Robustness to Word Substitution Attack with Differential Privacy

Wenjie Wang, Pengfei Tang, Jian Lou, Li Xiong


Abstract
The robustness and security of natural language processing (NLP) models are significantly important in real-world applications. In the context of text classification tasks, adversarial examples can be designed by substituting words with synonyms under certain semantic and syntactic constraints, such that a well-trained model will give a wrong prediction. Therefore, it is crucial to develop techniques to provide a rigorous and provable robustness guarantee against such attacks. In this paper, we propose WordDP to achieve certified robustness against word substitution at- tacks in text classification via differential privacy (DP). We establish the connection between DP and adversarial robustness for the first time in the text domain and propose a conceptual exponential mechanism-based algorithm to formally achieve the robustness. We further present a practical simulated exponential mechanism that has efficient inference with certified robustness. We not only provide a rigorous analytic derivation of the certified condition but also experimentally compare the utility of WordDP with existing defense algorithms. The results show that WordDP achieves higher accuracy and more than 30X efficiency improvement over the state-of-the-art certified robustness mechanism in typical text classification tasks.
Anthology ID:
2021.naacl-main.87
Volume:
Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies
Month:
June
Year:
2021
Address:
Online
Venue:
NAACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
1102–1112
Language:
URL:
https://aclanthology.org/2021.naacl-main.87
DOI:
10.18653/v1/2021.naacl-main.87
Bibkey:
Cite (ACL):
Wenjie Wang, Pengfei Tang, Jian Lou, and Li Xiong. 2021. Certified Robustness to Word Substitution Attack with Differential Privacy. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 1102–1112, Online. Association for Computational Linguistics.
Cite (Informal):
Certified Robustness to Word Substitution Attack with Differential Privacy (Wang et al., NAACL 2021)
Copy Citation:
PDF:
https://preview.aclanthology.org/emnlp-22-attachments/2021.naacl-main.87.pdf
Optional supplementary data:
 2021.naacl-main.87.OptionalSupplementaryData.pdf
Video:
 https://preview.aclanthology.org/emnlp-22-attachments/2021.naacl-main.87.mp4
Data
AG NewsIMDb Movie Reviews