Advancing Fake Profile Detection: A Machine Learning Perspective

The development of Internet has made individuals’ lives simpler and more convenient in many ways. People from different parts of the world communicate with one another through a variety of social media platforms that are accessible online. Users of these platforms contribute data in order to share significant moments and recollections from their lives. The quantity of data stored on these sites continues to increase, which brings with it a rise in the potential threats to users’ privacy. There have been several attempts by cybercriminal actors to hack into these websites and steal sensitive information. In order to accomplish this mission, attackers create duplicate user accounts, known as fake profiles. These fake profiles harm society by spreading fake news, rumors, stealing user personal information, defaming celebrities, asking for money for funds etc. So, identification of fake profiles is the most pressing need of society, in terms of the protection of private data. This paper discusses various machine learning algorithms used in fake profile detection by different researchers, and a comparison of these techniques with their results has been made for aiming at identifying fake profiles. In order to accomplish this task, two Instagram dataset, Dataset1 Instagram fake and real accounts dataset final-v1.csv, and Datase2 Instagram fake spammer genuine accounts dataset train.csv are used. The results obtained by both the datasets are compared using fifteen algorithms. The findings demonstrate that the result vary for same algorithm on different datasets. Voted Perceptron leads on dataset1, while SMO leads on dataset 2 in terms of Precision, Recall, F1-Score and Accuracy. These findings encourage us to conduct additional research in the related field.