Online Social Networks: An Efficient Framework for Fake Profiles Detection Using Optimizable Bagged Tree

Fake profiles pose a significant challenge to the integrity and security of online social networks (OSNs), making it imperative to develop reliable detection techniques. Spammers invariably adjust their strategies to circumvent filters, making it difficult to come up with impeccable design filters that can readily accommodate newly developed throes of fake profiles. Filters that primarily focus on textual information have limits while dealing with non-textual forms of spam, such as the ones based on images or audio. In this paper, we have employed a comprehensive set of features, including profile information, network characteristics, and behavioural patterns, to capture the distinguishing characteristics of fake profiles and conducted experiments on a large-scale dataset of OSN profiles, comprising both genuine and fake profiles. The Optimizable Bagged Tree algorithm allows us to reach optimize decision tree structure while leveraging the benefits of ensemble learning. By tenaciously pruning the tree’s structure and trimming irrelevant branches, the proposed framework achieves better generalization and robustness. The results demonstrated that our model outperforms traditional detection methods in terms of accuracy. Moreover, our approach exhibits high efficiency, enabling real-time detection of fake profiles in OSNs.