An Effect of Stacked CNN for Network Intrusion Detection System
Pankaj Rahi, Monika Dandotiya, A. Anushya, Ajay Khunteta, Pankaj Agarwal · 2022
The Intrusion Detection System (IDS) is a vital component of network security since it recognizes & inhibits hostile behavior. Because of the dynamic and time-varying nature of the network environment, network intrusion data are drowned out by a big number of normal samples, resulting in inadequate samples for model training & detection outcomes with a high percentage of false positives. An ID (Intrusion Detection) approach based on stacked Convolutional Neural Networks is discussed in this work to solve the issue of data imbalance. Modern network security requires more than conventional firewalls & data encryption technologies, which are no longer capable of meeting the demands of modern network security. It has as a consequence been advocated to use IDSs to cope with network threats. Recent mainstream ID methods are helped by Machine Learning and Deep Learning; however, they suffer from poor detection rates and the requirement for considerable feature engineering, which makes them less effective. This research presents DLNID (Deep Learning Model for Network Intrusion Detection) utilizing stacked CNN to improve detection accuracy. The experimental results demonstrate that this model's accuracy & F1- score by CNN on NSL-KDD and UNS are better than CNN.