stealthiness

5 papers

• A Set of Generalized Components to Achieve Effective Poison-only Clean-label Backdoor Attacks with Collaborative Sample Selection and Triggers
• Adjacent Words, Divergent Intents: Jailbreaking Large Language Models via Task Concurrency
• BAM-ICL: Causal Hijacking In-Context Learning with Budgeted Adversarial Manipulation
• SNEAKDOOR: Stealthy Backdoor Attacks against Distribution Matching-based Dataset Condensation
• Temporal Logic-Based Multi-Vehicle Backdoor Attacks against Offline RL Agents in End-to-end Autonomous Driving