Uncovering Fake Profiles Using Machine Learning Algorithms

Fake user profiles has been one of the biggest security issues in all online social networks (OSN) platforms. Whether operated by bots or humans, their sole presence disrupts the online social environment by being either a direct or indirect form of threat to users and platforms alike. Thus, detection of fake profiles plays a crucial role to mitigate this sort of issue, and with the help of machine learning algorithms this can be done simply and automatically while still achieving reliable and high-quality outcomes. In this paper, the research that was done employs and introduces an Ensemble Learning of three algorithms, namely Support Vector Machine (SVM), Random Forest algorithm, and Logistic Regression -- a combination that has been sparsely tested and trialed compared to their individual applications. The aim of this approach is to enhance the generalizability while maintaining robustness of the detection model. Synthetic Minority Oversampling Technique (SMOTE), a data preprocessing method was also applied before the training and testing to prevent risks from associated with imbalanced data. The model evaluation will be determined by obtaining the precision results of the conducted experiment, highlighting the proposed methodology's effectiveness in identifying fake accounts. The findings from the conducted study are expected to contribute to the improvement of automated methods, as well as to provide insights that may be beneficial in strengthening security and user trust across different OSN platforms.