DECEIVE-AFC：针对具有搜索功能的基于大语言模型的事实核查系统的对抗性声明攻击
DECEIVE-AFC: Adversarial Claim Attacks against Search-Enabled LLM-based Fact-Checking Systems
AI总结：
本文提出DECEIVE-AFC框架，通过新型声明级攻击策略和对抗性声明有效性评估原则，研究对抗性声明攻击对基于大语言模型的事实核查系统的影响，实验表明攻击显著降低验证性能。
摘要：
具有搜索功能的大语言模型（LLM）事实核查系统在动态检索外部证据以验证声明方面展现出强大潜力。然而，此类系统对对抗性攻击的鲁棒性仍不够清楚。本文在现实输入-only威胁模型下研究对抗性声明攻击，提出DECEIVE-AFC框架，整合新型声明级攻击策略和对抗性声明有效性评估原则。DECEIVE-AFC系统探索对抗性攻击轨迹，破坏搜索行为、证据检索和LLM推理，无需访问证据源或模型内部。在基准数据集和真实系统上的广泛评估表明，我们的攻击显著降低验证性能，将准确性从78.7%降至53.7%，并显著优于现有声明基攻击基线，具有强跨系统迁移性。
英文摘要：
Fact-checking systems with search-enabled large language models (LLMs) have shown strong potential for verifying claims by dynamically retrieving external evidence. However, the robustness of such systems against adversarial attack remains insufficiently understood. In this work, we study adversarial claim attacks against search-enabled LLM-based fact-checking systems under a realistic input-only threat model. We propose DECEIVE-AFC, an agent-based adversarial attack framework that integrates novel claim-level attack strategies and adversarial claim validity evaluation principles. DECEIVE-AFC systematically explores adversarial attack trajectories that disrupt search behavior, evidence retrieval, and LLM-based reasoning without relying on access to evidence sources or model internals. Extensive evaluations on benchmark datasets and real-world systems demonstrate that our attacks substantially degrade verification performance, reducing accuracy from 78.7% to 53.7%, and significantly outperform existing claim-based attack baselines with strong cross-system transferability.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
