Hybrid intrusion detection and signature generation using Deep Recurrent Neural Networks

Abstract

Automated signature generation for Intrusion Detection Systems (IDSs) for proactive security of networks is a promising area of research. An IDS monitors a system or activities of a network for detecting any policy violations or malicious actions and produces reports to the management system. Numerous solutions have been proposed by various researchers so far for intrusion detection in networks. However, the need to efficiently identifying any intrusion in the network is on the rise as the network attacks are increasing exponentially. This research work proposes a deep learning-based system for hybrid intrusion detection and signature generation of unknown web attacks referred as D-Sign. D-Sign is capable of successfully detecting and generating attack signatures with high accuracy, sensitivity and specificity. It has been for attack detection and signature generation of web-based attacks. D-Sign has reported significantly low False Positives and False Negatives. The experimental results demonstrated that the proposed system identifies the attacks proactively than other state-of-the-art approaches and generates signatures effectively thereby causing minimum damage due to network attacks.