MALWARE IMAGE PREDICTION AND CLASSIFICATION USING CONVOLUTIONAL NEURAL NETWORK

Abstract

Without users' permission, malware software can infect computers or other devices. Through these loopholes, criminals commit a range of illegal and criminal offences that violate the legitimate rights and interests of the nation. Traditional malware categorization techniques fall into two categories: static analysis and dynamic analysis. It is usually not necessary to execute malware binary samples in order to perform static analysis techniques, and disassembly makes it simple to recover important data such as text lists, routines, and hash values. The static analysis methods offer a high accuracy rate and a simple operation with a low consumption time. Static analysis tools, however, are limited to analyzing malware binary samples at the surface level, where they are readily influenced by deformation and other means of confusion. Furthermore, it is challenging to identify and categorize unknown malware. Methods of dynamic analysis are not impacted by obfuscation and can operate in a virtual environment. It has the ability to recognize newly discovered malware samples and track the dynamic alterations of malware binary samples over time. Nevertheless, it's an extremely intricate and time-consuming process. Malware has become one of the largest security threats in recent years due to its rapid growth. But feature engineering makes it difficult to handle large amounts of malware and readily limits the use of standard machine learning methods for malware categorization. However, dynamic analysis methodologies are not appropriate for efficiently categorizing large amounts of malware due to their complexity and high cost. In light of this, we propose a novel static malware detection method based on the convolutional neural network (CNN) employed in this work. Unlike existing methods, we use the data enhancement method to fix the unbalanced datasets, turn every viral byte into a colour image, and provide a better design.