| 引用本文: | 薛大暄,杜宜霏,陈红,李翠平.基于差分隐私的通信高效联邦推荐方法.软件学报,2026,37(6):2584-2606 |
| |
|
| |
|
|
| 本文已被：浏览 713次 下载 325次 |
码上扫一扫！ |
|
|
| 基于差分隐私的通信高效联邦推荐方法 |
|
薛大暄1,2,3, 杜宜霏1,2,3, 陈红1,2,3, 李翠平1,2,3
|
|
1.数据工程与知识工程教育部重点实验室 (中国人民大学), 北京 100872;2.数据库与商务智能教育部工程研究中心 (中国人民大学), 北京 100872;3.中国人民大学 信息学院, 北京 100872
|
|
| 摘要: |
| 推荐系统已成为大数据时代缓解信息过载问题的关键技术, 广泛应用于电子商务等领域, 但传统的集中式数据收集方式存在用户隐私泄漏的风险. 联邦学习允许多个数据持有者在不共享用户原始数据的情况下进行联合训练以保护数据隐私, 联邦推荐系统也受到工业界和学术界的广泛关注. 现有的联邦推荐算法将推荐系统的建模过程置于分布式环境中, 有效避免了用户敏感信息在中心服务器上的集中存储, 但仍存在隐私泄露和通信成本高的问题. 针对该问题, 提出一种基于差分隐私的通信高效联邦推荐算法. 该算法设计一种通用的子模型选择策略, 通过在客户端采用随机响应机制加强对用户交互数据的隐私保护, 并在服务器端采用最大似然估计的方法估计物品的真实交互频率来优化子模型的选择, 实现用户隐私保护与模型效用之间的有效平衡. 该算法不仅适用于矩阵分解推荐模型, 还可扩展应用于深度学习推荐模型, 在不同建模场景下均表现出较高的灵活性和适用性. 此外, 为进一步降低通信开销, 针对深度学习模型复杂结构和庞大参数导致的通信负担, 提出全局模型结构化划分策略, 并为浅层网络和深层网络制定差异化的优化策略, 有效降低了通信开销. 理论分析表明该方法满足差分隐私性质. 在真实数据集上的实验结果表明, 该方法在不显著降低模型可用性的前提下, 保障了用户数据的隐私安全, 同时大幅提高了联邦推荐中的通信效率. |
| 关键词: 联邦推荐 差分隐私 随机响应 深度学习 |
| DOI：10.13328/j.cnki.jos.007550 |
| 分类号:TP311 |
| 基金项目:国家重点研发计划(2023YFB4503600); 国家自然科学基金(U23A20299, U24B20144, 62172424, 62276270, 62322214) |
|
| Communication-efficient Federated Recommendation Method with Differential Privacy |
|
XUE Da-Xuan1,2,3, DU Yi-Fei1,2,3, CHEN Hong1,2,3, LI Cui-Ping1,2,3
|
|
1.Key Laboratory of Data Engineering and Knowledge Engineering of the Ministry of Education (Renmin University of China), Beijing 100872, China;2.Engineering Research Center for Database and Business Intelligence of the Ministry of Education (Renmin University of China), Beijing 100872, China;3.School of Information, Renmin University of China, Beijing 100872, China
|
| Abstract: |
| Recommendation systems have become a key technology in mitigating information overload in the era of big data, with widespread applications in E-commerce and other fields. However, traditional centralized data collection methods expose significant risks of user privacy leakage. Federated learning enables collaborative model training across multiple data holders without the need to share raw user data, thus protecting privacy. Federated recommendation systems have gained considerable attention from both academia and industry. Existing federated recommendation algorithms place the model training process in a distributed environment, effectively avoiding the centralized storage of sensitive user data on a single server. However, these approaches still face challenges related to privacy leakage and high communication costs. To address these issues, this study proposes a communication-efficient federated recommendation algorithm based on differential privacy. The algorithm introduces a general sub-model selection strategy that strengthens privacy protection of user interaction data on the client side through a randomized response mechanism. On the server side, it employs maximum likelihood estimation to infer the true interaction frequencies of items and optimize the sub-model selection process. This strategy achieves an effective balance between privacy protection and model utility. The proposed algorithm is applicable not only to matrix factorization-based recommendation models but also to deep learning-based models, demonstrating high flexibility and adaptability across various recommendation scenarios. Furthermore, to reduce communication overhead, a global model partitioning strategy is proposed to address the complex structures and large parameter sizes of deep learning models. Differentiated optimization strategies are applied to shallow and deep networks to effectively mitigate communication costs. Theoretical analysis shows that the method satisfies differential privacy, while experimental results on real-world datasets demonstrate that the proposed approach preserves user data privacy without significantly compromising model utility, while substantially improving communication efficiency in federated recommendation systems. |
| Key words: federated recommendation differential privacy (DP) randomized response deep learning |
|
|
|
|
