Recommendation system based on localized differential privacy

Abstract

The invention discloses a recommendation system based on localized differential privacy, and the main process comprises the following steps that: (1) each user side encodes related historical commodity purchasing behavior data, then disturbs the data by using a random disturbance algorithm meeting the localized differential privacy, and finally sends the disturbed data to a server side; (2) the server side collects all disturbed data, then reconstructs relevant information of historical purchase behaviors of the user, and finally maps the relevant information into a low-dimensional vector by using a graph embedding algorithm and returns the low-dimensional vector to the user side; (3) each user side calculates a respective recommended candidate set through the finally generated vector, thereby finishing commodity recommendation under the condition of privacy protection. According to the invention, a localized differential privacy protection technology is used, and the system does not need a trusted third party as a medium, thereby guaranteeing that the privacy information of each user is not stolen by the third party, and also guaranteeing the usefulness of a final recommendation result of the system.

Description

technical field [0001] The invention relates to a recommendation system for privacy protection, in particular to a recommendation system based on localized differential privacy. Background technique [0002] Local Differential Privacy (Local Differential Privacy) is a privacy protection technology that does not require a trusted third party and can be applied to the protection of individual privacy data in data analysis. In localized differential privacy, each individual who needs privacy protection uses a random perturbation algorithm that satisfies the localized differential privacy mechanism to perturb their own data; data analysts analyze the relevant statistics of the original individual data by collecting perturbed data feature. Localized differential privacy provides parameters for controlling the strength of privacy, through which the security and usability of the random perturbation algorithm can be balanced. [0003] Graph Embedding (Graph Embedding) is to map th...

AI Technical Summary

Problems solved by technology

The recommendation system using encryption technology requires complex encryption and decryption calculations on the client and server sides.

These systems undoubtedly still have problems in terms of security and efficiency in the absence of a trusted third party

Images