Hybridmsbd: Malicious Social Bot Detection Using Hybrid Learning Automata-Based Malicious Social Bot Detection

Abstract

Because of SM (Social Media), people are more likely to share their expertise on the internet. They may also communicate with individuals from all around the globe thanks to it. With all its attention, it has turned into a fun place to try out new attacks. User data (e.g. user-produced data) in OSNs has grown significantly in terms of volume, velocity, and variety. New techniques for gathering and analyzing such comprehensive data have been investigated as a result. By automating analytical services and improving customer service quality, SBs have proven useful. Fake news (i.e. misinformation) has had real-world repercussions for MSB, on the other hand. As a result, it is essential to identify and remove MSB (Malicious Social Bots) from OSNs the vast majority of existing MSB identification methods are quantitative in nature. Consequently, the analytical accuracy is low when using SBs since they easily imitate these features. HybridMSBD, a new method based on hybrid learning automata, was proposed for the identification of MSBs. KNN, SVM, and Naive Bayes, Random Forest and Decision Tree algorithms, were utilized as pre-processors for the ML models. A novel method is presented that combines behavior detection utilizing the HLA-MSBD algorithm to identify tweets produced by genuine individuals or bots. The first simulation findings are very promising. Experiments on the Kaggle dataset indicate that our method can outperform current state-of-the-art bot identification algorithms in performance.