A Hybrid Approach with Intrinsic Feature-Based Android Malware Detection Using LDA and Machine Learning

World is becoming small with the increase in the number of mobile phone users. The most influential and having huge market among mobile phones is android. Android is a software used in nowadays smart phones, which not only consists of operating system but also myriad number of key applications. These applications make large number of day to day tasks easy. There are millions of android applications in the market with over 3 billion or more downloads. The growing market of this platform not only invites smart phone users, but it also becomes a point of interest for black hat hackers. Hackers use this technology for large number of activities by spreading the android applications in this platform which are not actually android packages rather malicious codes or malware. Therefore, these malwares must be handled in a smart way; otherwise, they lead to huge loss. Different techniques have been used for detection of android malware which consists of network traffic analysis, static analysis, and dynamic analysis. In this paper, a combined approach of static, dynamic, and intrinsic features for android malware detection using k-nearest neighbor (k-NN), random forest, decision tree, SVM, and ensemble learning techniques. The calculation uses a publicly available dataset of Androtrack. The estimation results shows that both the decision tree and random forest classifiers produced accuracy of 99%. With the help of newly added feature and a different approach of preprocessing, i.e., linear discriminant analysis.