Design and Analysis of Contemporary Neural Architectures for Network Intrusion Detection System

Abstract

newline With the rapid evolution of Internet of Things (IoT) network advancement has caused a huge influence on the increasing count of devices and advanced enhancements linked with it. This raises the influence of emerging various cyber attacks and this causes the necessity to design a robust security application. Therefore, it is a quintessential necessity to establish an efficient security system that identifies the attacks, known as Intrusion Detection System (IDS). Owing to the network complexities in IoT, conventional Machine Learning (ML) algorithms possess certain limitations while processing a huge volume of data. Such limitations can be overcome by the Deep Learning (DL) techniques that have the benefit of refining numerous data features. With this incredible advantage, DL method is highly desirable for malware detection and categorization. This article devises a unified solution for network intrusion detection utilizing hybrid deep-learning based optimized model. The first contribution is developed for mitigating the network intrusion detections using multimodal networks. The crucial advantage of this designed approach is that it highly strengthens the individual features better when compared to the normalization technique. In the second contribution, an effective optimization-enabled deep learning network is exploited for identifying the malwares in the network. Here, malicious breaches are identified employing Deep Maxout Network (DMN), where weights of the classifier are optimally adjusted using Remora Optimization Algorithm (ROA). The third contribution is to propose a robust technique for intrusion detection exploiting hybrid optimization-based hybrid deep learning model named Remora Whale Optimization (RWO). Here, RV coefficient is utilized for accomplishing the feature selection mechanism, wherein the hybrid deep learning model like DMN and Deep autoencoder are employed for network intrusions. Moreover, hybrid deep learning technique is devised by incorporating ROA and Whale Optimization A