A Hierarchical Attention-Based Neural Network Model for Socialbot Detection in OSN

In this paper, we present a hierarchical attention-based deep neural network model for socialbots detection in online social networks through modeling user behavior at two levels of granularity leveraging profile, activity, temporal, and content information. The proposed approach is novel from two perspectives - (i) it models user representations using a comprehensive set of profile and behavior information, and (ii) it applies hierarchical attention mechanism at both low-level and high-level user representations. The proposed approach jointly models profile, temporal, and activity information as sequences, which are given as input to a two-layer stacked BiLSTM, while content information is given to a six-layer CNN. On evaluation, our model performs significantly better in comparison to the baselines and state-of-the-art methods.