A Novel Cyber Security Malware Detection Using Optimization Based Fine-Tuning Dual-Channel Convolutional Neural Network (DCCNN)With Spider Monkey Optimization (SMO) And Hierarchical Particle Swarm Optimization (HPSO)

Abstract

The implementation of conventional IoTs (Internet of Things) concepts to industrial sectors as well as apps is known as the Industrial Internet of Things (IIoT). Smart cities, smart grids, linked automobiles, smart homes, and supply chain management are just a few areas where IIoT is being used. Cybercriminals, however, are increasingly focusing on those networks. It has been demonstrated that conventional static techniques for IoT malware detection and evaluation are inadequate for comprehending the behaviour of IoT malware in order to avoid and mitigate its effects. The fields of DL (Deep Learning) and BDA (Big Data Analytics) have enormous promise for creating strong security protocols for IIoT networks. This research presents a hybrid DL strategy to identify files contaminated with malware on an IoT network.In order to detect malware, a hybrid DCCNN (Dual-Channel Convolutional Neural Network) named Hybrid SMO-HPSO, which combines Spider Monkey Optimisation (SMO) and Hierarchical Particle Swarm Optimisation (HPSO) is suggested. CB-STM-RENet is a DCCNN Fine-Tuning (FT) technique that uses the STM concept in a novel way. To address this, researchers have put forth a novel convolutional block STM that can perform region-basededge-based and region-basedprocedures both independently and in tandem. Investigating intensity inhomogeneity, boundary-defining features, and region homogeneity is made easier by the methodical application of edge and region algorithms in conjunction with convolutional processes. I.T. The visual representation of unpacked binary files for both malicious and safe apps is the malware dataset utilised for this investigation.This dataset is a result of the IoT malware detection Google Code Jam (GCJ). According to the outcomes of the research, the suggested solution outperforms current approaches by Acc, R, specificity, P, MCC, F1-score, and an average time for processing for each malware classification when it comes to measuring cyber security risks in the IoTs.