A Recommendation Algorithm Based on Matrix Completion Against Torture Attack

Abstract

The invention discloses a shilling attack resistance recommendation algorithm based on matrix completion. The algorithm comprises the following steps that: firstly, carrying out statistics on the scoring of a user for an item, extracting the attribute characteristics of the user and the attribute characteristics of the item, and independently constructing a user-item scoring matrix, a user attribute characteristic matrix and an item attribute characteristic matrix; then, modeling shilling attack scoring in the user-item scoring matrix into structured noise in a matrix completion model; then, adopting a blocking coordinate descent algorithm to carry out iterative update on each variable, and solving a structured line noise matrix; subsequently, according to the structured line noise matrix,rejecting the shilling attack scoring in the user-item scoring matrix; and finally, using a traditional recommendation algorithm to carry out scoring prediction, and solving a prediction scoring matrix. By use of the recommendation algorithm provided by the invention, the shilling attack user in a recommendation system can be effectively detected, a personalized scoring prediction effect which ismore accurate than a traditional recommendation algorithm can be obtained under shilling attack interference, and the robustness of the recommendation algorithm is effectively improved.

Description

technical field [0001] The invention belongs to the field of information security in the field of computer technology, and in particular relates to an anti-trust attack recommendation algorithm based on matrix completion. Background technique [0002] Facing the problem of information overload, the recommendation system came into being. A recommendation system is a software system that understands user preferences by collecting user information, item information, and user-item interaction information, thereby recommending items that users may be interested in to users, and to a certain extent solves the information that troubles users. overload problem. One of the mainstream algorithms currently used to implement recommendation systems is the collaborative filtering algorithm. It relies on users' historical behavior, analyzes past user-item interactions, and establishes new user-item connections. However, the generator of user-item interaction data is all users, and no ac...

AI Technical Summary

Problems solved by technology

[0005] Purpose of the invention: Based on the structured noise matrix completion technology, the present invention proposes a robust personalized recommendation algorithm against trolling attacks to improve the ability of the recommendation system to resist trolling attacks. First, detect trolling attacking users in the recommendation system and eliminate corresponding attack scores Then recommend, so as to solve the problem of trust attack in the recommendation system, and recommend suitable items to the corresponding users

Images