n-gram Effect in Malware Detection Using Multilayer Perceptron (MLP)

Abstract

Malware is a threat that can compromise cyber security. Currently, the development of malware is becoming increasingly complex and difficult to detect. One way to improve detection accuracy is to implement the n-gram feature extraction. n-gram is one of method to analyze malware, by capturing the frequency of string/opcode which often appear from malware. This work aims to improve the performance of malware detection by evaluating the best number of n-grams to extract the opcode. Selection of n number in n-gram process will be very influencing in malware classification result. This research work investigates the effect the n value of n-gram on the accuracy detection by varying the value n = 1 to n = 5. The best accuracy detection in the experiments using Multilayer Perceptron (MLP) classifier reaches 89 percent.