A novel machine learning and face recognition technique for fake accounts detection system on cyber social networks

Online Social Networks (OSN) such as Facebook, Instagram, Twitter, and others have seen rapid growth in recent years. Such applications provide attractive online social networks and communications with the opportunity to connect with relatives and acquaintances, meet new people, enter communities, talk, exchange photos, organize events, and network with others who are close to real-life; unfortunately, on the other hand, they also raise privacy and security issues. We identified OSN threats in this paper and recommended a digital face-processing authentication method as a double-factor authentication after entering the password using Matlab. After applying deep learning classification by attending to a real dataset from the live webcam to train the model, we achieved the best accuracy rate of 95%. However, such methods have yet to be deployed to all social networks, so we also mentioned the problem of fake accounts, which is one of the most significant problems in OSN. These are effective tools for executing spam campaigns and spreading malware and phishing attacks. Fake accounts could lead to the loss of money for businesses, loss of reputation, stealing information for malicious purposes, and much more. This study is related to detecting fake and legitimate profiles on OSN. For this purpose, we chose two datasets that contain fake and legitimate accounts on Facebook and Instagram. Each contains different features after applying machine learning using Naive Bayes, Logistic Regression, Support Vector Machines, K-Nearest Neighbour, Boosted Tree, Neural Networks, SVM Kernal, and Logistec Regression Kernal. SVM achieved the highest classification accuracy for the Fake Profiles detection datasets with 97.1%.