ROBUST INTELLIGENT MALWARE DETECTION USING DEEP LEARNING

Abstract

Vulnerabilities caused by malware attacks continue to increase and have become a significant security problem in the digital age. Malware detection is still a hot research topic because malware attacks are increasing and affecting many computer users, businesses, and governments. Currently, malware is seeking solutions in static and dynamic analysis of malware signatures and behavioral patterns; This is very time consuming and has proven to be ineffective at instantly detecting known malware. Recent malware uses polymorphism, morphing, and other evasion methods to rapidly change malware behavior and create a flood of new malware. This new type of malware is often different from existing malware, and machine learning algorithms (MLA) have recently been adopted for effective malware analysis. However, such processes take a long time because complex architecture requires special training and artistic representation. The engineering process can be completely avoided by using advanced MLA methods such as deep learning. Recently published studies in this direction have shown the performance of their algorithms with biased data, which actually limits the realtime use of their strategies. There is an urgent need to reduce bias and selfevaluate this process to find new ways to detect badly dated malware. To fill this gap in the literature, this paper first evaluates classical MLA and deep learning for the detection, classification and classification of malware using different public and private databases. Second, by making a distinction between public and private data and using different time periods to report and evaluate the standard in different ways, we eliminate any data biases removed from the test analysis. Third, our main contribution will be to propose a new image processing method with the negative view of MLA and deep learning to achieve effective malware detection models. A comparative study of our models shows that our deep learning architecture outperforms classical MLA. We are delivering the first zeroenergy day in malware detection by providing visualization and deep learning for integrated static, dynamic and image processing applications in big data environments. Overall, this article presents a method for effective malware detection using deep learning for rapid deployment. Static and dynamic analysis, artificial intelligence, machine learning, deep learning, image processing, scalable and hybrid frameworks. Introductionn the digital world of Industry 4.0, rapid advances in technology affect your daily business activities and personal life. Internet of Things (IoT) and ApplicationsSenior Researcher Corrado Mencar participated in the review of this text and approved its publication.