Dynamic graph neural network for fake news detection

Abstract

The widespread of fake news on social media and other platforms can bring significant damage to the harmony and stability of our society. To defend against fake news, researchers have suggested various ways of dealing with fake news. In recent years, fake news detection has become the research focus in both academic and industrial communities. The majority of existing propagation-based fake news detection algorithms are based on static networks and they assume the whole information propagation network structure is readily available before performing fake news detection algorithms. However, real-world information diffusion networks are dynamic as new nodes joining the network and new edges being created. To address these shortcomings, we proposed a dynamic propagation graph-based fake news detection method to capture the missing dynamic propagation information in static networks and classify fake news. Specifically, the proposed method models each news propagation graph as a series of graph snapshots recorded at discrete time steps. We evaluate our approach on three real-world benchmark datasets, and the experimental results demonstrate the effectiveness of the proposed model.