DAMAGE: Detecting Adversarially Modified AI Generated Text
This addresses the challenge of detecting adversarially modified AI text for security and integrity applications, presenting a novel solution with practical implications.
The paper tackled the problem of AI-generated text being modified by humanizer tools to evade detection, showing that many existing detectors fail against such text, and demonstrated a robust detector that maintains low false positive rates and cross-humanizer generalization.
AI humanizers are a new class of online software tools meant to paraphrase and rewrite AI-generated text in a way that allows them to evade AI detection software. We study 19 AI humanizer and paraphrasing tools and qualitatively assess their effects and faithfulness in preserving the meaning of the original text. We show that many existing AI detectors fail to detect humanized text. Finally, we demonstrate a robust model that can detect humanized AI text while maintaining a low false positive rate using a data-centric augmentation approach. We attack our own detector, training our own fine-tuned model optimized against our detector's predictions, and show that our detector's cross-humanizer generalization is sufficient to remain robust to this attack.