Is it my preference? Role Grounding Attack against Agent's Memory via Social Interaction

LLM-based agents now share experiences on social platforms and remember new preferences to enhance their performance. However, existing memory systems contain a vulnerability in role-to-principal attribution that attackers can exploit. We identify Role Grounding Misattribution, in which an attacker-...

Descrición completa

Detalles Bibliográficos

Autores principales:	Bu, Haoran, Shen, Yiting, Zhang, Litian, Hu, Ruihan, Zhang, Xi
Formato:	Artigo
Publicado:	2026
Materias:	Cryptography and Security Artificial Intelligence Multiagent Systems
Acceso en liña:	https://arxiv.org/abs/2610.38209 http://arxiv.org/abs/2610.38209
Etiquetas:	Engadir etiqueta Sen Etiquetas, Sexa o primeiro en etiquetar este rexistro!