Machine Learning Based Detection of Fake Accounts on Online Social Media: A Feature Engineering

Abstract

The prevalence of online social networks (OSNs) among today’s youth has led to a close connection between social activities and these websites. However, the growth of OSNs and the user data on these platforms has also attracted attackers and imposters who engage in detrimental activities like stealing user data, spreading false information, and creating fake accounts. Individual account analysis and coordinated activity detection are two types of techniques that scholars have developed for identifying bogus accounts and suspicious behavior to address these issues. The article proposes a feature engineering method to efficiently detect fraudulent social media profiles and bots by using dimension reduction, feature selection, and data preprocessing techniques. Additionally, support vector machines, neural networks, Ada Boost Classifiers, random forests, and decision trees are used as machine learning categorization methods. The project’s goals are to make fake social media accounts easier to spot and shed light on the role that false identities play in sophisticated persistent threats.