Enhancing Intrusion Detection using Deep Learning and An Improved Conditional Variational AutoEncoder (ICVAE)

Abstract

In the age of the internet and its vast user base, numerous attacks are launched every day. The requirement for information network security has grown significantly over the past few years due to the enormous expansion of Internet applications. An intrusion detection system is intended to be able to adjust to the constantly shifting threat landscape as the main defense of network infrastructure. A particular branch of machine learning called "deep learning" uses structures resembling neurons to learn new information. By making enormous strides in a variety of fields, including speech processing, computer vision, and natural language processing, to mention a few, deep learning has fundamentally altered the means by which we approach learning tasks. Effective outcomes in intrusion detection systems are demonstrated by machine learning techniques. We present ICVAE-DNN, a novel intrusion detection model that combines an Improved Conditional Variational AutoEncoder (ICVAE) with a deep neural network (DNN). An effective model was trained by optimizing unsupervised SAE. According to the experimental findings, the suggested model outperforms conventional approaches in terms of total rate of detection and low percentage of false positives. The KDDCup99, NSL-KDD, and UNSW-NB15 datasets were used to evaluate the suggested model. Using the UNSW-NB15 dataset, the model achieved a 99% accuracy rate and a 97.99% detection rate.