Adversarial Style Augmentation via Large Language Model for Robust Fake News Detection
This addresses the critical social challenge of fake news spread for individuals and society, though it is incremental as it builds on existing detection methods to enhance robustness against attacks.
The study tackled the problem of fake news detection being vulnerable to style-conversion attacks by proposing AdStyle, an adversarial style augmentation method using LLMs to generate diverse attack prompts, resulting in significantly improved robustness and detection performance on benchmark datasets.
The spread of fake news harms individuals and presents a critical social challenge that must be addressed. Although numerous algorithmic and insightful features have been developed to detect fake news, many of these features can be manipulated with style-conversion attacks, especially with the emergence of advanced language models, making it more difficult to differentiate from genuine news. This study proposes adversarial style augmentation, AdStyle, designed to train a fake news detector that remains robust against various style-conversion attacks. The primary mechanism involves the strategic use of LLMs to automatically generate a diverse and coherent array of style-conversion attack prompts, enhancing the generation of particularly challenging prompts for the detector. Experiments indicate that our augmentation strategy significantly improves robustness and detection performance when evaluated on fake news benchmark datasets.