Genuine or Spammer? Enhanced Fake Profile Detection using Feature Synthesis

Social networks are now an inalienable part of our lives. Originally created for communication, these sites have become an essential tool in business, learning, traveling, and almost every other part of life. With increasing usage, the interaction has been growing at a very fast pace, which has attracted the wrong people to use these platforms for various vices. One of the most common techniques used includes the development of fake accounts, which enable the offenders to engage in malicious activities while remaining anonymous. This is a very dangerous situation for users and their security and safety. In order to solve this issue, it is important to detect fake profiles at an early stage. In this paper, we present a detailed framework for detecting fake profiles on social media platforms. The framework consists of three key feature engineering techniques: feature pre-processing, feature extraction namely Additive-based Feature Synthesis, and feature selection for profile characteristics. Using a public dataset of 696 Instagram profiles, the proposed framework was evaluated to measure its effectiveness in fake profile detection. The findings show that the proposed framework is efficient with the Gradient Boosting Classifier at an accuracy of 93.26% and the Ensemble model at an accuracy of 93.12%. Also, the F1 scores of these models were 0.9306 and 0.9272 respectively indicating that the models have a good recall-precision trade-off. The comparison with three other models showed that the proposed framework provides higher accuracy and F1-score each time. These results indicate that the proposed framework can be a practical and accurate method to identify fake profiles on social networks.