Research on Privacy Protection of Recommendation System Users Under Member Inference Attack Based on IMIA Defender Neural Network

With artificial intelligence entering a new era, recommendation systems are entering a stage of excellent development. It is necessary to propose a recommendation proposition that better meets the user's privacy protection needs, focusing on security thinking centered on user privacy. Using the dynamic evolution of the development of Federated Recommendation System (FedRecs), a theoretical analysis framework is developed based on the internal logic of privacy protection, which explains the development mechanism of FedRecs that is generated jointly by the federated learning mechanisms and privacy defense mechanisms involving users and service providers. Further, explores the possibility of moving towards the high-quality development goal of FedRecs from the perspective of privacy attack and defense changes and practical interpretations of FedTRecs. The purpose of the development of FedRecs is to provide users with recommendation services that meet expected standards and are committed to continuously improving recommendation quality and enhancing user satisfaction. For this purpose, it is necessary to strengthen privacy control based on the interaction-level member inference attack (IMIA) defender neural network, develop an interaction and feedback mechanism between FedRecs and user privacy perception, and establish a system for evaluating FedRecs and user privacy perception. In this way, the high-quality development of FedRecs can be realized, artificial intelligence security can be promoted, and user privacy can be truly satisfied.