Search papers, labs, and topics across Lattice.
6
0
7
21
Linking software vulnerabilities to attacker behaviors reveals critical insights for proactive threat mitigation in cybersecurity.
Expert specialization in malware detection leads to a remarkable 97.44% accuracy, even against adversarial mutations.
Current safety mechanisms for LLMs are critically flawed, as a genetic algorithm can effectively evolve adversarial prompts to bypass them entirely.
Forget retraining: NeWTral instantly restores safety to your LLM after adding a risky LoRA, slashing attack success rates from 70% to 13% without sacrificing expertise.
Certifiable defenses against malware evasion are now possible without modifying the underlying ML architecture, offering a practical path to robust security.
LLM-as-a-Judge, while improving evaluation scalability, is riddled with unexplored security risks, making it a potential target and instrument for attacks.
