CleanNews: a Network-aware Fake News Mitigation Architecture for Social Media

Abstract: With the widespread use of the internet and handheld devices, social media now holds a power similar to that of old newspapers. People use social media platforms for quick and accessible information. However, this convenience comes with a variety of risks. Anyone can freely post content, true or false, with the probability of remaining online forever. This makes it crucial to identify and tackle misinformation and disinformation on online platforms. In this article, we propose CleanNews, a comprehensive architecture to identify fake news in real-time accurately. CleanNews uses advanced deep learning architectures, combining convolutional and bidirectional recurrent neural networks, i.e., LSTM and GRU, layers to detect fake news. A key contribution of our work is a novel embedding technique that fuses textual information with user network structure, allowing the model to jointly learn linguistic and relational cues associated with misinformation. Furthermore, we use two network immunization algorithms, i.e., SparseShield and NetShield, to mitigate the spread of false information within networks. We conduct extensive ablation studies to evaluate the contribution of each model component and systematically tune hyperparameters to maximize performance. The experimental evaluation on two real-world datasets shows the efficacy of CleanNews in combating the spread of fake news.