Detection of Evolutionary Malware Using Machine Learning on Large Scale Data

Abstract

According to recent theoretical and empirical research, malware is one of the biggest threats facing by the digital world. In order to maintain security, malware mitigation strategies have developed throughout time. In the past, a number of traditional techniques were employed to identify malware that included elements like heuristics and signatures, among others. The complex obfuscation techniques employed by new generations of malware were too formidable for conventional malware detection systems to handle. Creating a malware detection system these days is difficult, particularly when dealing with malware that is of a new generation. New malware generations have been able to evolve thanks to advanced evasion techniques, which has had extremely serious consequences. Malware attacks on computer systems typically occur over the Internet, namely via email, harmful websites, and drives that have associated software code. For a variety of purposes, including financial gain, stealing private or secret data, turning the system into a bot or rendering services unavailable, etc., computer systems are hacked. Static analysis encompasses the examination of malware samples without their execution. Features are derived directly from the code or file structure, enabling the identification of probable malware attributes without the need to execute harmful code, therefore minimizing the associated risks. The Dynamic malware identification technique involves executing executable files in a controlled environment to watch and analyze their behavior. This facilitates the capture of the immediate actions and consequences of the virus, which may not be apparent through static analysis. In this study the malware dataset is being collected from Virustotal, Kegel, VXheaven. A total of 29,710 malware files collection for hybrid analysis. Both strategies aim to extract distinctive information that can distinguish between malicious and benign files. The efficacy of the malware detection system primarily relies on the accurate identificati