A Survey on Machine Learning to Detect Creation of Fake Identities by Human vs. Bots

Online social networks are more prevalent than ever and have become deeply ingrained in people’s social lives. Through online social networks, they chat with one another, share data, organize events, and even run their own online companies. In order to steal personal information, spread destructive activities, and publish fake information, attackers and imposters have been drawn to OSNs because of their explosive growth and the vast amount of personal data they collect from their users. On the contrary, researchers have started to look into reliable strategies for identifying fake accounts and questionable activities using account attributes. However, several of the employed account variables have no impact at all or have a negative effect on the results. Furthermore, employing independent categorization algorithms does not necessarily yield positive outcomes. In order to effectively detect phone Instagram accounts, a novel algorithm called SVM-NN is proposed in this research. There were utilized four feature evaluation and data reduction procedures. The support vector machine, neural network, and the most current technique, SVMNN, were used to determine whether the chosen accounts were actual or spam. SVM-NN outperforms SVM and NN, using fewer characteristics while still being able to accurately classify about 89% of the users in our classification dataset.