A novel method for malware detection using audio signal processing techniques

Abstract

Malware is any kind of program that is designed to perform malicious activity in computers and networks. To evade traditional signature-based malware detection techniques, malware developers employ obfuscation techniques. Two main type of obfuscation techniques are polymorphism and metamorphism. New approaches for detecting obfuscated malwares are commonly based on machine learning techniques to extract features from binary bytes, opcode sequence or function calls of programs. In this paper, we propose a novel method based on audio signal processing techniques. We represent binary bytes as audio signals and then apply Music information retrieval (MIR) techniques to find musical patterns using MFCC and chromagram features in them. Then, we create machine learning model based on extracted features in order to classify new samples. Our technique is light weight and consumes low memory. To evaluate effectiveness of our method, we have done several experiments. The results of these experiments show that our technique has good accuracy while it is low memory and time consumption.