Shilling attack detection method and system based on dynamic knowledge graph, equipment and medium

Abstract

The invention provides a shilling attack detection method based on a dynamic knowledge graph. The shilling attack detection method comprises the following steps: collecting user information; calculating personal preferences and social styles of the users; establishing a user relationship dynamic knowledge graph in combination with the change of the user relationship along with the dynamic change factor; and according to the condition that the user relationship of the real user changes along with the dynamic change factor and the user relationship of the shilling attack user does not change along with the dynamic change factor after the propagation capability goes through, carrying out user clustering on the user relationship dynamic knowledge graph by utilizing a graph group clustering method so as to detect the shilling attack user. According to the shilling attack detection method based on the dynamic knowledge graph, the user is simulated as a social sensor, the user relationship dynamic knowledge graph is established in combination with the change of the user relationship along with the dynamic change factor, and the dynamic change of the user relationship in the social activity is accurately described; the shilling attack users are analyzed and detected by gathering the user relationships which change along with the dynamic change factors at a plurality of moments, and the shilling attack users can be detected more accurately.

Description

technical field [0001] The present invention relates to the technical field of trolling attack detection, in particular to a trolling attack detection method, system, device and computer-readable storage medium based on a dynamic knowledge graph. Background technique [0002] With the continuous development of Internet technology, especially the explosive growth of news, commodities and entertainment resources, people also face a serious problem of information overload when they turn to the Internet to search for the items they need. In order to provide people with information quickly and accurately, recommender systems are widely used in various fields. Due to the openness and other characteristics of the recommendation system itself, it gives the opportunity for users to inject false ratings and false social relationships into the system. Among them, the attacking user changes the real user's social relationship by establishing a large number of false relationships with o...

AI Technical Summary

Problems solved by technology

Supervised detection methods mainly detect troll attacks by training classifiers, such as extracting features from user registration information, user posting content and other information to build classifiers. Supervised detection methods require a large number of samples, and manual labeling of samples is costly; The unsupervised detection method mainly uses the topological relationship of the social network to identify abnormal points in the network, such as the detection method of clustering posts by using the similarity between text and URL, etc. The unsupervised detection method has a high false positive rate and low robustness. slightly inferior; while the semi-supervised detection methods that combine the two mostly start from the similarity between users, user ratings, etc., and rarely consider the dynamic relationship between users

Images