Deep Learning for Robust Intelligent Malware Detection

Abstract

“Malware or malware remains a major concern in present digital life as
computer users, companies and governments see the rise of malware attacks. Current
malware detection solutions are use static and dynamic analysis of malware signatures and
behaviours; this is time consuming and ineffective in identifying unknown malware. The
latest malware uses polymorphism, shapeshifting, and other avoidance methods to quickly
change the behavior of the malware and create more malware. Recently, machine learning
algorithms (MLA) have been used to effectively identify malware because new malware is
often different from existing malware.” This requires a lot of engineering skills, technical
training and artistic expression. The engineering process can be avoided altogether by using
advanced MLA techniques such as deep learning. Although some recent studies have
progressed in this direction, the efficiency of algorithms is highlighted by the data.
“In order for newly developed methods to be effective in zero-day malware detection, it is
necessary to reduce the bias and self-testing of these methods. To fill this gap in the
literature, this study evaluates the role of classical MLA and deep learning in malware
detection, classification and classification using public and private databases. Training and
testing distinguish between public and private data used in clinical trials and collected at
different times. We also provide a new image processing system with state-of-the-art MLA
and in-depth courses.” A qualitative analysis of the method shows that deep learning
outperforms traditional MLA. Overall, this project uses deep learning techniques to classify
malware in real time and provides powerful insights. Visualization and deep learning, based
on a combination of static, dynamic and image processing in a big data environment, is a
new method for zero-day malware discovery.