Node status anti-attack method based on graph neural network in directed symbol network

Abstract

The invention discloses a node status confrontation attack method based on a graph neural network in a directed symbol network. The method comprises the following steps: step 1, constructing a social status evaluation model of nodes in the directed symbol network; and 2, constructing a directed symbol network-oriented node status adversarial attack model, and modifying a graph according to attack requirements based on the social status evaluation model constructed in the step 1 to obtain an adversarial attack graph. According to the method, the research of the status theory is introduced and combined aiming at the signed directed graph, and the previous related research does not aim at graph data; in addition to different data types of the graph, the stability of other nodes except the target node is considered in the constraint of designing the attack target function, the status evaluation result of the target node is improved on the basis of maintaining the stability of other nodes, and a test proves that compared with similar algorithms, the method has the advantages that the reliability is high, and the reliability is high. The method is better in attack performance.

Description

technical field [0001] The invention relates to the field of social networks, in particular to a node position confrontation attack method based on a graph neural network in a directed symbolic network. Background technique [0002] In recent years, with the vigorous development of online social platforms, the interaction between online users has become traceable, making it possible to study the relationship between users in online social networks, and graph neural networks emerge as the times require. As an emerging graph data learning technology, graph neural network realizes the deep integration of graph data and deep learning, and has received extensive attention from academia and industry. The research and application of graph neural network has been extended to the fields of node classification, link prediction, community detection, and research and development of drug molecules. [0003] The influence maximization problem has always been a research hotspot in social ...

AI Technical Summary

Problems solved by technology

One-hop nodes (neighbors), two-hop nodes, and even N-hop nodes of the target node all have an impact on the status of the target node, but the influence is different. How to find the best node is one of the difficulties; 2) Deleting or adding edges will Change the structure of the graph, but the social network will affect the whole body, even if only one edge is changed, the status of several nodes and even the status of the entire network node will be affected

Images