SilverSpeak: Evading AI-Generated Text Detectors using Homoglyphs

TL;DR

This paper demonstrates that homoglyph-based attacks can effectively bypass multiple state-of-the-art AI-generated text detectors across various datasets, highlighting vulnerabilities in current detection methods.

Contribution

The paper introduces homoglyph-based attack techniques and evaluates their effectiveness against seven detectors, revealing significant vulnerabilities in existing AI-generated text detection methods.

Findings

Homoglyph attacks reduce detector accuracy significantly.

All tested detectors are vulnerable to homoglyph-based circumvention.

Detection performance drops from an average MCC of 0.64 to -0.01.

Abstract

The advent of Large Language Models (LLMs) has enabled the generation of text that increasingly exhibits human-like characteristics. As the detection of such content is of significant importance, substantial research has been conducted with the objective of developing reliable AI-generated text detectors. These detectors have demonstrated promising results on test data, but recent research has revealed that they can be circumvented by employing different techniques. In this paper, we present homoglyph-based attacks (A $\rightarrow$ Cyrillic A) as a means of circumventing existing detectors. We conduct a comprehensive evaluation to assess the effectiveness of these attacks on seven detectors, including ArguGPT, Binoculars, DetectGPT, Fast-DetectGPT, Ghostbuster, OpenAI's detector, and watermarking techniques, on five different datasets. Our findings demonstrate that homoglyph-based attacks can effectively circumvent state-of-the-art detectors, leading them to classify all texts as either AI-generated or human-written (decreasing the average Matthews Correlation Coefficient from 0.64 to -0.01). Through further examination, we extract the technical justification underlying the success of the attacks, which varies across detectors. Finally, we discuss the implications of these findings and potential defenses against such attacks.