A Reinforcement Learning-Based Graph Adversarial Example Generation Method by Adding False Nodes for Document Classification

Abstract

The invention discloses a graph adversarial sample generation method by adding false nodes based on reinforcement learning, including: (1) obtaining the original graph number and graph node classification model, constructing a training set and a test set; adding false nodes to the original graph data , get the initial adversarial samples; (2) construct the attack model; (3) select the attack target in the training set; (4) input the current adversarial samples and attack targets into the attack model, select the node with the largest evaluation value, and construct a new adversarial sample; (5) input the new adversarial sample into the classification model, if the classification result is the target result, get the adversarial sample and proceed to the next step, otherwise skip to step (4); (6) train the attack model, and use the training Good attack models to test and apply. The invention generates graph adversarial samples by adding false nodes, which can help to design a more robust graph deep learning model.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence information security, and in particular relates to a method for generating graph confrontation samples by adding false nodes based on reinforcement learning. Background technique [0002] A graph in graph theory is a graph composed of a number of given points and a line connecting two points. This graph is usually used to describe a certain relationship between certain things. Points represent things, and points connect two points. The line of represents that there is a certain relationship between the corresponding two things. The graph G in graph theory is an ordered pair (V, E), where V is called the vertex set, which is the set of all vertices in the graph, and E is called the edge set, which is the set of edges between all vertices. gather. Simply put, vertices represent things, and edges represent relationships between things. In addition, the attribute graph (Attributed G...

AI Technical Summary

Problems solved by technology

In the field of graph data, current research is to make the classification results of target nodes mislead by adding or deleting existing edge or node features.

But this method may be difficult to implement in actual scenarios. For example, in a social network, if you want to delete or add an edge between two users, you may need to obtain the login permissions of these users, but this is difficult to obtain in actual situations.

Images