T-Bot: AI-based social media bot detection model for trend-centric twitter network

Online social media platforms have paved the way greatly into the lives of people and made person-to-person interaction faster by making the world more connected. Users can spread their content across a vast audience worldwide. Significantly, the microblogs have paved their way and have become crucial grounds for public discussions, marketing, and political campaigns. Due to the increase in social media usage, the automated profiles with deviant behaviour started flooding the online social media, thus creating a cybersecurity concern. These automated accounts are also known as social media bots. The proliferation of such malicious automated bots in social networks has increased the structural complexity, thereby making bot detection a considerable cybersecurity threat. Though, there exist certain mechanisms for the detection of bots, but most of them rely only on the basic user profile attributes. Our research presents a novel artificial intelligence (AI)-driven bot detection framework, called T-Bot to find the automated bot accounts in Twitter network using basic as well as derived attributes. The proposed T-Bot facilitates bot detection in trend-centric datasets using an innovated centroid initialization algorithm (CIA), especially to handle unbalanced datasets. Our proposed CIA performs a binary classification of our initial dataset that is used for feature selection and construction of ruling hierarchy. The parameters of machine learning models are hyper-tuned to achieve optimality in deriving automation score to the suspected bots in trend-centric social networks.