Which Neurons Detect Malicious Code? A Probing Study of LLM Security Knowledge

🗓️ 11 Jul 2026 00:00:00Reported by Vang T. Nguyen, Phuong T. Nguyen, Anh M. T. Bui, Lam D. DaoType

packetstormnews

packetstormnews🔗 packetstorm.news👁 8 Views

This study locates FFN neurons responsible for malware detection in three instruction tuned LLMs using causal interventions.

11 Jul 2026 00:00Current

6.2Medium risk

Vulners AI Score6.2

malware detection neuron interpretability large language models feed forward network causal interventions model behavior security aware alignment code security