Cyber-attack detection and identification using deep learning

Abstract

The quantity and complexity of cyber-attacks are both on the rise. For defenses to keep up with the ever-evolving threats, it will need ever-greater technological advances and fresh ideas. Traditional security methods like intrusion detection and deep packet inspection are still used and recommended, but they are not enough to keep up with the growing number of security threats. The widespread usage of APT's communication networks makes them more susceptible to assaults by cybercriminals, with possibly catastrophic outcomes. Critical infrastructures are monitored and controlled by Advanced Persistent Attack using real-time monitoring to identify aberrant behaviors of the system. The most existing Advanced Persistent Threat defenses were created to protect IT infrastructure and are seldom useful in more robust settings like factories. The primary objective of this thesis is to use various learning-based approaches to evaluate network traffic and sensory measures in real-time in order to identify and locate cyber-attacks. In order to achieve this, numerous learning-based models are presented, such as a self-tuning and scalable deep learning and classification model for cyber-attack site identification and an ensemble deep learning-based cyber-attack detection method for unbalanced Advanced Persistent Threat datasets. Two real-world advanced persistent threat datasets are used to assess the effectiveness of the suggested models. In terms of f1-score, recall, and accuracy, the models presented do better than the most recent research.