Scholarlink

<p>As the volume of digital information continues to grow, recommender systems have evolved into indispensable mechanisms for extracting and presenting content aligned with users’ specific interests and needs. However, they face critical challenges, including data silos, privacy concerns, scalability limitations, and the difficulty of handling distributed and heterogeneous data sources. Federated recommender systems (FRS) have emerged as a promising paradigm, enabling collaborative model training while preserving user privacy by keeping raw data decentralized. Despite their potential, FRS still encounter open challenges such as data heterogeneity, high communication costs, cold-start issues, personalization trade-offs, and security and privacy constraints, which remain active areas of research. This survey provides a comprehensive and systematic review of the FRS literature, offering a structured analysis of architectural designs, learning paradigms, personalization strategies, communication mechanisms, and privacy-preserving techniques. In addition, we include an experimental section that provides quantitative analyses of performance, communication costs, and execution time across representative studies. By synthesizing both qualitative and quantitative insights, this survey not only organizes the current state of the art but also identifies actionable directions for future research in federated recommender systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Privacy meets personalization: a systematic literature review of federated recommender systems

  • 期刊论文

  • 发表日期：

    2026年4月30日

查看全文
  • Marwa Badrouni,
  • Wissem Inoubli,
  • Chaker Katar,
  • Zahra Kodia

摘要

As the volume of digital information continues to grow, recommender systems have evolved into indispensable mechanisms for extracting and presenting content aligned with users’ specific interests and needs. However, they face critical challenges, including data silos, privacy concerns, scalability limitations, and the difficulty of handling distributed and heterogeneous data sources. Federated recommender systems (FRS) have emerged as a promising paradigm, enabling collaborative model training while preserving user privacy by keeping raw data decentralized. Despite their potential, FRS still encounter open challenges such as data heterogeneity, high communication costs, cold-start issues, personalization trade-offs, and security and privacy constraints, which remain active areas of research. This survey provides a comprehensive and systematic review of the FRS literature, offering a structured analysis of architectural designs, learning paradigms, personalization strategies, communication mechanisms, and privacy-preserving techniques. In addition, we include an experimental section that provides quantitative analyses of performance, communication costs, and execution time across representative studies. By synthesizing both qualitative and quantitative insights, this survey not only organizes the current state of the art but also identifies actionable directions for future research in federated recommender systems.