A State-of-the-Art Technique for Malware Detection Based on Deep Learning in IoT Environment

Abstract

Internet of Things (IoT) devices are used to sense, gather, transmit and control data. IoT improves user experiences by connecting and sharing information across smart devices. Malware attacks easily target IoT-connected consoles to get control or damage privacy. Effective IoT device protection might rescue millions of internet users from malware. Malware detection methods are computationally expensive and complex. Many different machine learning models were used to deduct malware or viruses. As innovation advances, threat intelligence and malware researchers struggle. Traditional security measures like firewalls and Intrusion Detection Systems (IDS) must be changed to work with the current IoT models. Also, the Internet of Things and Cloud/Fog Computing go well together. They are often used interchangeably when talking about technical services and work together to make IoT services more complete. This study suggests deep learning-based approaches for detecting IoT malware in order to prevent harm to IoT devices. The proposed algorithms AlexNet, ResNet-18, and Convolution Neural Network (CNN) are applied to the well-known public dataset "Mal_Img". The experimental findings show that the accuracy of the suggested method is superior to that of earlier state-of-the-art and deep learning methods. The highest attained accuracies for proposed methods proposed algorithms AlexNet, ResNet-18, and Convolution Neural Network are 97.74%, 88.74%, and 98.02%. Our framework has currently been demonstrated to be a useful research tool for the BitDefender AntiMalware Research Labs' computer security specialists. In the future, more advanced malware (polymorphic and metamorphic) can be detected using the same models.