AuthorMist: Evading AI Text Detectors with Reinforcement Learning

2025-03-10 · Isaac David, Arthur Gervais

In the age of powerful AI-generated text, automatic detectors have emerged to identify machine-written content. This poses a threat to author privacy and freedom, as text authored with AI assistance may be unfairly flagged. We propose AuthorMist, a novel reinforcement learning-based system to transform AI-generated text into human-like writing. AuthorMist leverages a 3-billion-parameter language model as a backbone, fine-tuned with Group Relative Policy Optimization (GPRO) to paraphrase text in a way that evades AI detectors. Our framework establishes a generic approach where external detector APIs (GPTZero, WinstonAI, Originality.ai, etc.) serve as reward functions within the reinforcement learning loop, enabling the model to systematically learn outputs that these detectors are less likely to classify as AI-generated. This API-as-reward methodology can be applied broadly to optimize text against any detector with an accessible interface. Experiments on multiple datasets and detectors demonstrate that AuthorMist effectively reduces the detectability of AI-generated text while preserving the original meaning. Our evaluation shows attack success rates ranging from 78.6% to 96.2% against individual detectors, significantly outperforming baseline paraphrasing methods. AuthorMist maintains high semantic similarity (above 0.94) with the original text while successfully evading detection. These results highlight limitations in current AI text detection technologies and raise questions about the sustainability of the detection-evasion arms race.

📄 PDF Abstract BibTeX arXiv:2503.08716

Code (0)

등록된 구현이 없습니다.

Tasks

reinforcement-learning Reinforcement Learning Semantic Similarity Semantic Textual Similarity Text Detection

TH-Bench: Evaluating Evading Attacks via Humanizing AI Text on Machine-Generated Text Detectors

2025-03-10 · Jingyi Zheng, Junfeng Wang, Zhen Sun, Wenhan Dong 외

As Large Language Models (LLMs) advance, Machine-Generated Texts (MGTs) have become increasingly fluent, high-quality, and informative. Existing wide-range MGT detectors are designed to identify MGTs to prevent the sprea…

Misinformation

Evading Deep Learning-Based Malware Detectors via Obfuscation: A Deep Reinforcement Learning Approach

2024-02-04 · Brian Etter, James Lee Hu, Mohammedreza Ebrahimi, Weifeng Li 외

Adversarial Malware Generation (AMG), the gen- eration of adversarial malware variants to strengthen Deep Learning (DL)-based malware detectors has emerged as a crucial tool in the development of proactive cyberdefense. …

Deep Reinforcement Learning Malware Detection reinforcement-learning Reinforcement Learning+1

StealthRL: Reinforcement Learning Paraphrase Attacks for Multi-Detector Evasion of AI-Text Detectors

2026-02-09 · Suraj Ranganath, Atharv Ramesh arxiv

AI-text detectors face a critical robustness challenge: adversarial paraphrasing attacks that preserve semantics while evading detection. We introduce StealthRL, a reinforcement learning framework that stress-tests detec…

Reinforcement Learning Text Detection

AdaDoS: Adaptive DoS Attack via Deep Adversarial Reinforcement Learning in SDN

2025-10-23 · Wei Shao, Yuhao Wang, Rongguang He, Muhammad Ejaz Ahmed 외 arxiv

Existing defence mechanisms have demonstrated significant effectiveness in mitigating rule-based Denial-of-Service (DoS) attacks, leveraging predefined signatures and static heuristics to identify and block malicious tra…

Reinforcement Learning

REPLICANT: Learning Policies for Evading and Hardening Malware Detectors

2026-08-28 · Shae McFadden, Ilias Tsingenopoulos, Mario D'Onghia, Alexander Herzog 외 arxiv

To determine the real-world effectiveness of machine learning based malware detection, it is vital to evaluate its robustness against highly capable adversaries. However, state-of-the-art attacks do not effectively model…

Reinforcement Learning Malware Detection