Twitter Bot detection using Graph based Deep Learning (Graph Neural Networks)

The concept of openness defines Twitter's ecosystem, facilitating automated account management through its API, resulting in the proliferation of ''bots''.These bots, while copying human actions such as tweeting and following, also engage in detrimental activities like disseminating false information and malicious software.Detecting & addressing these bots is imperative to combat misinformation & cyber threats.Our research introduces a pioneering approach utilizing Graph Neural Networks (GNNs) to harness Twitter's network structure for precise bot identification.We establish a dynamic graph model of the Twittersphere, representing user interactions as edges to comprehensively grasp relational dynamics.By enriching this graph with usercentric data including account longevity's & activity's, we obtain a holistic understanding.Our tailored GNN architecture, drawing inspiration from prior models, learns user embedding to discern patterns indicative of bot behavior's, thereby revealing the covert domain of automated accounts.Through the integration of network analysis, machine learning, and social media analytics, our interdisciplinary methodology aims to contribute significantly towards mitigating the detrimental effects of bot-driven misInformation on Twitter.