Personalized search system for enhancing privacy protection based on federated learning

Abstract

Through a method in the field of artificial intelligence, the personalized search system for enhancing privacy protection based on federated learning is achieved, a hardware architecture of the systemis composed of a client and a server, a personalized search framework based on federated learning is constructed, a specifically trained underlying model is a personalized sorting model, and the personalized sorting model is a personalized search framework based on federated learning. Clients and data stored in the clients jointly participate in training of a personalized sorting model in a federated learning mode, the trained model is deployed on each client, query is initiated on the clients, search history H of a user is stored, a user portrait P is constructed, non-personalized results returned from a server is rearranged, and the rearranged non-personalized results are displayed to the user. The problem of protecting the privacy of the user when the user interest is mined by utilizing the query history of the user to deduce the current query intention is solved; based on the framework, two models of FedPSFlat and FedPSProxy are designed, so that the problem of data heterogeneityis solved, and the problems of performance bottleneck, communication obstacle and privacy attack faced by single-layer FedPSFlat are solved.

Description

technical field [0001] The invention relates to the field of artificial intelligence intelligent search, in particular to a personalized search system based on federated learning to enhance privacy protection. Background technique [0002] Personalized search is mainly to adjust the document list based on user interests to better meet different query intentions expressed by different users using the same ambiguous query. Existing related work mainly includes: traditional personalized search models based on topics, clicks or other features and personalized search models based on deep learning. These models need to use personal information such as users' historical query sequences and click behaviors to infer user interests and specific query intentions, so there is a risk of leaking user privacy. [0003] The current privacy protection technology in search mainly considers the identifiability and linkability of privacy. Recognizability refers to identifying who the user is,...

AI Technical Summary

Problems solved by technology

However, utilizing only a single user's personal data is not enough to train a reliable personalized search model

Images