Enhancing Fake News Detection Using Graph Autoencoders and LSTMs

The proliferation of fake news has emerged as a critical challenge in today's information age, necessitating robust methods for its detection and mitigation. In this paper, we propose a comprehensive approach for detecting fake news in social networks by integrating graph neural networks, specifically graph autoencoders (GAE), with Long Short-Term Memory (LSTM) networks. Initially, propagation graphs are constructed from news articles, which are then fed into graph autoencoders to calculate graph embeddings. Subsequently, to capture the temporal patterns of news propagation, an LSTM layer is employed on the embeddings. The final output of the LSTM is processed through a simple multi-layer perceptron to predict the authenticity of the news. Experimental results demonstrate the effectiveness of our approach, achieving superior performance in detecting fake news compared to existing methods. This study underscores the potential of combining structural and temporal features for fake news detection in online social networks.