AI-Powered Security Information and Event Management: A Review of Deep Learning Approaches for Modern Cyber Security

Cyber threats are increasing with rapidly emerging digital infrastructure. In that scenar-io, even a reasonable pace cannot be maintained by traditional SIEM systems. Deep learn-ing technologies such as Variational Autoencoders and Graph Neural Networks have to be embedded into SIEM based systems so that anomaly activity might be detected and zero-day threats also identified with accurate and timely generation of the alerts. We critically compare recent advances in deep learning-based SIEM with their more practical applica-tions in detecting complex cyberattacks and focus on how AI improves the effectiveness of the efficacy of reports from SIEMs and technical challenges associated with those method-ologies.