Are AI-Generated Text Detectors Robust to Adversarial Perturbations?

TL;DR

This paper introduces SCRN, a new AI-generated text detector that is more robust to adversarial attacks by using a reconstruction network and siamese calibration, significantly improving detection accuracy and generalizability.

Contribution

The paper proposes the Siamese Calibrated Reconstruction Network (SCRN), a novel method that enhances robustness of AI-generated text detection against adversarial perturbations.

Findings

SCRN outperforms baseline methods by 6.5%-18.25% accuracy under attacks.

SCRN demonstrates superior cross-domain and cross-genre robustness.

The method improves detection reliability against adversarial noise.

Abstract

The widespread use of large language models (LLMs) has sparked concerns about the potential misuse of AI-generated text, as these models can produce content that closely resembles human-generated text. Current detectors for AI-generated text (AIGT) lack robustness against adversarial perturbations, with even minor changes in characters or words causing a reversal in distinguishing between human-created and AI-generated text. This paper investigates the robustness of existing AIGT detection methods and introduces a novel detector, the Siamese Calibrated Reconstruction Network (SCRN). The SCRN employs a reconstruction network to add and remove noise from text, extracting a semantic representation that is robust to local perturbations. We also propose a siamese calibration technique to train the model to make equally confidence predictions under different noise, which improves the model's robustness against adversarial perturbations. Experiments on four publicly available datasets show that the SCRN outperforms all baseline methods, achieving 6.5\%-18.25\% absolute accuracy improvement over the best baseline method under adversarial attacks. Moreover, it exhibits superior generalizability in cross-domain, cross-genre, and mixed-source scenarios. The code is available at \url{https://github.com/CarlanLark/Robust-AIGC-Detector}.