A trusted recommendation scheme for privacy protection based on federated learning
- PDF / 1,476,052 Bytes
- 11 Pages / 595.276 x 790.866 pts Page_size
- 80 Downloads / 259 Views
REGULAR PAPER
A trusted recommendation scheme for privacy protection based on federated learning Yichuan Wang1 · Yuying Tian1 · Xinyue Yin1 · Xinhong Hei1 Received: 15 June 2020 / Accepted: 31 October 2020 / Published online: 23 November 2020 © China Computer Federation (CCF) 2020
Abstract With the convergence of the era of global news and the era of big data, the daily amount of news sent to the world is exploding. Users also face the problem of information overloads when they get massive information, which leads to how cloud servers push personalized data to users among massive data have become the focus of news companies. In order to obtain the push accuracy, the traditional recommendation system often makes deep mining of users’ privacy data, which makes users’ privacy cannot be guaranteed. In order to solve the above problems, this paper proposes a collaborative filtering algorithm recommendation system based on federated learning on end-edge-cloud. The exposure of data privacy was further prevented by adding Laplace noise to the training model through differential privacy technology. Finally, the training model and recommendation information is stored to the blockchain network to provide permanent storage, evidence chain and real-time traceability services.On the premise of protecting data privacy, this system provides cloud server with solutions to alleviate computing pressure, bandwidth pressure and improve news push accuracy through end-edge-cloud distributed learning. Keywords Federated learning · Blockchain · Differential privacy · Recommendation system
1 Introduction The emergence of personalized recommendation technology is to solve the problem of accurate distribution of information and help users find the information they want. Personalized recommendation technology plays an important role in reducing user cost, improving work efficiency, optimizing content and improving background performance. Recommender system is a kind of information filtering tool, which can present the most relevant content of interest to specific users by using the user characteristics and habits of the whole cluster. With the explosive growth of information on the Internet, it is impossible for users to browse all the content. The recommendation system can help users find the Electronic supplementary material The online version of this article (https://doi.org/10.1007/s42045-020-00045-8) contains supplementary material, which is available to authorized users. * Xinhong Hei [email protected] Yichuan Wang [email protected] 1
Xi’an University of Technology, Xi’an, China
13
Vol:.(1234567890)
real interesting content according to their taste by analyzing the data. However, the recommendation system is based on the centralized processing mode of cloud computing, which inevitably consumes a lot of communication resources and causes a large time delay in the data transmission process; moreover, the physical limited communication resources (such as bandwidth) will limit the scalability of the system, The transmissio
Data Loading...
