DESIGN AND IMPLEMENTATION OF HYBRID METHOD TO IMPROVE ANDROID MALWARE DETECTION ACCURACY BASED ON ONLINE ENSEMBLE MACHINE LEARNING

• Main Author: Azizah, Nida
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/54511
• Institution: Institut Teknologi Bandung

Android is an operating system that is widely used by public. The development of Android and the number of applications in it causes malware take a change to grow and develop in these applications, so a malware detection method is needed on Android. Malware detections are needed so that applications infected by malware do not enter the Android marketplace and malware will not spread out. Machine learning had become one of detection method that use to detect malware in Android. However, malware is evolving rapidly causing detection accuracy using machine learning to decrease. The solution to this problem is to use online learning as a malware detection method on Android. Online learning is a part of machine learning. Several studies have succeeded in solving the problem of malware detection on Android using online learning and even increasing accuracy. However, a new problem emerged, namely these studies using static analysis at the feature extraction stage. Static analysis is one of feature analysis technique. This technique has drawback such as susceptibility to code obfuscation, dynamic code loading and has low affectivity and accuracy. Therefore, it is necessary to conduct research aimed at increasing accuracy and overcoming the problems caused by the implementation of static analysis. In this study, a framework will be designed to improve the accuracy of malware detection on Android in terms of analysis techniques as well as to overcome problems due to the implementation of static analysis on malware detection on Android using online learning.