SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2402.11638 · ACL · 2024

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

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

arXiv · PDF · Open in the Atlas

Code that ran

We have not lifted any functions out of this paper's repositories yet, so there is nothing we have run. The repositories linked to it are listed below.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

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 machinegenerated text detectors under attacks from diverse categories: editing, paraphrasing, cogenerating, 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. 1 Attack Category Method Model-Free? Level Access Detailed Descriptions Editing ( §6.2) post-generation Typo Insertion ✓ Character None Create typos by inserting, deleting, substituting, and transposing mainly. Homoglyph Alteration ✓ Character None Change English characters into visually similar Unicodes, e.g., Cyrillic characters. Format Character Editing ✓ Character None Change or insert formatting characters, including zero-width whitespace \u200B insertion, and shift character editing, e.g., \n, \r, \u000B (vertical tab), etc.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2402.11638")
get_code_for_paper("2402.11638")
have("2402.11638")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2402.11638.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2402.11638.svg)](https://syntology.ai/paper/2402.11638)

Connect an agent — have() is free.