Machine Learning Based Malicious Detection Approaches: A Survey

Abstract

Malware is harmful software designed to breach user privacy, endanger computer systems, or obtain unauthorized access to networks. As a result of the growing number of uses for computers and the reliance on electronically stored private information, malware attacks on confidential data are becoming a serious problem for people and businesses worldwide. Thus, malware protection is essential to maintaining the security of our personal computing systems and data. Recent research articles have alternately concentrated on single-attacking techniques or malware detection systems. As far as we are aware, no survey article presents malware patterns of attack and protection techniques in tandem. This work attempts to tackle this problem by combining machine learning (ML) oriented models of detection for complex and contemporary malware with a variety of malicious attack methods. This allows us to concentrate on a taxonomy of malware attacks according to four basic dimensions: the attack's main objective, its mode of attack, its targeted exposure as well as execution process, and the kinds of malware that carry out the attack. Thorough details on methods for analysing malware are also looked into. Furthermore, a thorough discussion is held regarding current malware detection methods that use ML algorithms and feature extraction.