Malware Detection in Mobile Devices Using Hard Voting Ensemble Technique

The steady rise in devices based on the Android operating system has ignited malware developers to invest their energy in launching various attacks targeting these devices. These attacks have caused pandemonium in various spheres of life including data breaches, displaying irritating graphics and loss of money. This has consequently sparked a need to develop techniques that aim at detecting these bad pieces of code in the devices. Lately, various tools that are aimed at detecting these bad pieces of software have been crafted using various machine learning algorithms. This paper reviews some of the work that has been done so far in creating malware detection tools using these algorithms. It then presents a proposed model that is aimed at performing malware detection on these devices using a unique way of incorporating the hard voting ensemble technique, combining Random Forest, Decision Tree, and Support Vector Machine techniques rather than using individual algorithms. This approach, coupled with the latest and very well-balanced dataset comprising 29298 malware files and 29298 benign files resulted in a very efficient model. The technical novelty of our proposed model includes the use of hard voting ensemble classifiers as well as using a large and well-balanced dataset with the latest malware application samples. It was discovered that the hard voting method outclassed all other individual classifiers in terms of accuracy and false positive rates. The model managed to perform malware detection with a satisfying 98.78
^%
accuracy as well as a 0.28% false positive rate.