Abstract

The proliferation of social media platforms has increased the risk of cyber-attacks, necessitating advanced techniques for timely detection and prevention. This study investigates the application of hybrid Machine Learning (ML) and Deep Learning (DL) models for social media attack prediction and detection. By integrating ML’s ability to extract salient features with DL’s strength in modeling complex temporal and spatial patterns, hybrid models demonstrate enhanced accuracy and robustness in identifying malicious activities. The research examines various architectures, including CNNs, LSTMs, GRUs, and autoencoders, highlighting their effectiveness in handling high-dimensional, imbalanced, and dynamic social media data. The findings underscore the potential of ML/DL hybrid frameworks to provide real-time, adaptive, and reliable solutions for mitigating cyber threats, ultimately contributing to safer online communication environments.