LATIDIA · Ciberseguridad

Ataques automatizados de inferencia de membresía (AutoMIA): descubrimiento de cálculos de señales MIA utilizando agentes LLM

arXiv: 2603.19375v2Tipo de anuncio: reemplazar Resumen: Los ataques de inferencia de membresía (mias), que permiten a los adversarios determinar si puntos de datos específicos formaban parte del conjunto de datos de entrenamiento de un modelo, han surgido como un impo

La noticia

arXiv:2603.19375v2 Announce Type: replace Abstract: Membership inference attacks (MIAs), which enable adversaries to determine whether specific data points were part of a model's training dataset, have emerged as an important framework to understand, assess, and quantify the potential information leakage associated with machine learning systems. Designing effective MIAs is a challenging task that usually requires extensive manual exploration of model behaviors to identify potential vulnerabilities. In this paper, we introduce AutoMIA -- a novel framework that leverages large language model (LLM) agents to automate the design and implementation of new MIA signal computations. By utilizing LLM agents, we can systematically explore a vast space of potential attack strategies, enabling the