Information Retrieval, Cybersecurity, adversarial risks, AI security evaluation, empirical experiments, performance evaluation, large language model security, trustworthy AI, trustworthiness evaluation, privacy-preserving AI, secure information processing, ablation analysis, benchmark construction, query rewriting, privacy leakage, dataset preparation