Fake Profile Detection on Online Social Networks Using Machine Learning

Abstract

Online Social Networks (OSNs) have become indispensable part for modern communication, commercial activity, and the dissemination of information. Platforms such as Twitter enable users to interact, share opinions, and engage with content at an unprecedented scale. However, this widespread adoption has also made OSNs attractive targets for malicious actors. Among the most pressing security concerns are profile cloning and fake engagement tactics that involve the creation and use of fake or automated accounts to manipulate popularity metrics, deceive genuine users, and distort public perception. These activities not only pose serious threats to individual privacy but also undermine trust in digital interactions, skew advertising strategies, and compromise the overall integrity of the platform. In this study, we focus on detecting such malicious behaviors on Twitter through the application of various machine learning algorithms. By conducting extensive experiments and performance evaluations, we compare traditional machine learning techniques with deep learning models. Our results demonstrate that deep learning approaches, particularly those leveraging neural network architectures, significantly outperform classical methods, achieving higher accuracy in identifying fake profiles and offering more robust solutions for OSN security.