Identifying Fake Account in Facebook Using Machine Learning

Abstract

Nowadays people rely vigorously on online social networks (OSNs) that have attracted cyber criminals’ interest in performing malicious acts. Furthermore, with the existence of illicit businesses that provide transactions of fake account services. This study focuses on identifying fake accounts in Facebook which is the most widely used in OSN. The methodology of this study is started with data collection, features identification and learning classifiers. The first process is to collect information on true and fake Facebook accounts. The second process is the use of Facebook user feed data to comprehend user profile activity and to identify a comprehensive collection of 5 characteristics that play a critical role in discriminating against fake users with true users on Facebook. Lastly, we use these characteristics and the identification of main classifiers based on machine learning that perform well in the assignment of identification out of a total of 3 classifiers namely K-nearest neighbour (KNN), support vector machine (SVM) and neural network (NN). The result shows that KNN generate 82% of the highest performing classifiers with classification precision. The findings have revealed that “likes” and “remarks” add well to the job of detection. However, although the precision value is not highly perfect, the findings of this study shows that most fake accounts are able to imitate actual users.