[go: up one dir, main page]

Stumbling Blocks: Stress Testing the Robustness of Machine-Generated Text Detectors Under Attacks

Yichen Wang, Shangbin Feng, Abe Hou, Xiao Pu, Chao Shen, Xiaoming Liu, Yulia Tsvetkov, Tianxing He


Abstract
The widespread use of large language models (LLMs) is increasing the demand for methods that detect machine-generated text to prevent misuse. The goal of our study is to stress test the detectors’ robustness to malicious attacks under realistic scenarios. We comprehensively study the robustness of popular machine-generated text detectors under attacks from diverse categories: editing, paraphrasing, co-generating, and prompting. Our attacks assume limited access to the generator LLMs, and we compare the performance of detectors on different attacks under different budget levels. Our experiments reveal that almost none of the existing detectors remain robust under all the attacks, and all detectors exhibit different loopholes. Averaging all detectors, the performance drops by 35% across all attacks. Further, we investigate the reasons behind these defects and propose initial out-of-the-box patches.
Anthology ID:
2024.acl-long.160
Volume:
Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Month:
August
Year:
2024
Address:
Bangkok, Thailand
Editors:
Lun-Wei Ku, Andre Martins, Vivek Srikumar
Venue:
ACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
2894–2925
Language:
URL:
https://aclanthology.org/2024.acl-long.160
DOI:
10.18653/v1/2024.acl-long.160
Bibkey:
Cite (ACL):
Yichen Wang, Shangbin Feng, Abe Hou, Xiao Pu, Chao Shen, Xiaoming Liu, Yulia Tsvetkov, and Tianxing He. 2024. Stumbling Blocks: Stress Testing the Robustness of Machine-Generated Text Detectors Under Attacks. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2894–2925, Bangkok, Thailand. Association for Computational Linguistics.
Cite (Informal):
Stumbling Blocks: Stress Testing the Robustness of Machine-Generated Text Detectors Under Attacks (Wang et al., ACL 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.acl-long.160.pdf