Abstract

Internet of Things (IoT) is an evolving digital technology, which is mainly meant to bridge physical and virtual world. New business model has been emerged because of people, objects, machines and Internet connectivity along with new interactions amid humanity and remaining world. IoT is considered as a gateway for cyber-attacks since various resources such as systems, applications, data storage, and services are connected through IoT that relentlessly provide services in the organization. IoT security is challenging factor due to prevailing software piracy and malware attacks presently. The economic and reputational damages are caused by these threats due to crucial information burglary. IoT malware detection is yet another challenging factor due to security design deficiency besides IoT devices specific characteristics such as processor architecture heterogeneity, particularly on identifying cross-architecture IoT malware. Hence, IoT malware detection area is main objective of this research by security community recently. The familiar dynamic or static analyses to detect IoT malware is greatly deployed in various researches with its benefits. A systematic review relating to latest research studies and technologies of classical, Deep Learning (DL) and Machine Learning (ML) methodologies for cyber security threats recognition are outlined and are view is given in this paper. Every approach pertaining to its objective, approach and outcomes have been examined for every selected work. Deep Learning (DL) approach is greatly utilized for malware infected files and pirated software recognition in IoT network in cloud. Several software piracy and malware attack detection methods has been analyzed in this paper with respect to its advantages and disadvantages. The source code plagiarism is detected through DL methodology and dataset collection is done from Google Code Jam (GCJ) for software piracy investigation. Rather than this, Deep Convolutional Neural Network (DCNN) is mainly involved in identifying malicious infections in IoT network. Mailing dataset is utilized for obtaining malware samples which is used for experimental purpose. It is thereby substantiated that suggested method namely Tensor Flow Deep Neural Network (TF-DNN) classification performance for assessing cyber security threats in IoT are enhanced when compared with classic approaches such as Support Vector Machine (LBP+SVM), Gray Level Cooccurrence Matrix with Support Vector Machine (GLCM+SVM) pertaining to F-measure (F1) and Classification Accuracy (CA).