Enhancing Fake Account Detection on Facebook Using Boruta Algorithm

With the increasing prevalence of social media platforms, detecting fake accounts has become a critical challenge for maintaining online security and user trust. This paper proposes a feature selection approach for detecting fake accounts on Facebook, aiming to improve the efficiency and accuracy of existing detection models. The research focuses on identifying a subset of informative features that contribute significantly to distinguishing between genuine and fake accounts. Thanks to Boruta algorithm, machine learning classifiers can yield similar detection performance to using all features (non-selection), potentially leading to computational and resource savings. In addition, this feature selection algorithm allows faster training and prediction times. The approach’s effectiveness is evaluated using a real Facebook dataset, and the results demonstrate its ability to improve interpretability and keep only the significant features.