Analysis of Hybrid Deep Learning Models for Efficient Intrusion Detection

Enhanced with digital capabilities over the preceding ten years, especially on the internet, there has been a surge in cyber attackers attempting to profit from consumers’ private information. One of the most recent approaches to solving this problem is to devise a model to help us detect acts of intrusion in a network known as Intrusion Detection Systems (IDS). There has been a surge in the development of advanced artificial algorithms in recent years, and adding their intelligence makes it much easier to prevent network intrusions. In general, Deep learning methods outperform machine learning techniques in terms of performance metrics for detecting network intrusions. This paper suggests the implementation of five distinct variants of deep learning models that have been extensively trained with the UNSW NB-15 benchmark dataset by optimizing different hyperparameters to provide improved intrusion detection. Amongst variants, the Bi-Directional Long-Short Term Memory model with ADAM optimizer is efficient in intrusion detection with the highest accuracy of 95.28%.