A Review of Deep Learning Approaches for Fake Profile Detection on Social Networking Sites

Social networking sites, now with thousands in existence, have ushered in a revolution in digital communication, while also giving rise to serious security threats like fake profile creation and online impersonations. The perpetrators engaged in these deceitful acts use them for cyberbullying, spreading misinformation, and identity theft, among other evils. Traditional detection methods relying on rule-based systems and shallow, machine learning algorithms have had modest success at best against the increasing complexity of fake profiles. Deep learning approaches have emerged in recent years as powerful, complementary alternatives capable of modeling highly complex patterns from large-scale, heterogeneous sources of data. This review paper presents an in-depth evaluation of state-of-the-art deep learning architectures such as Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Autoencoders, and Transformer-based architecture in the domain of fake profile detection. The paper also looks at multimodal methods combining textual, image, behavioral, and network-based features to enhance detection accuracy. Challenges tackled herein include class imbalance, data privacy, adversarial evasion, and real-time implementation.