A Robust Android Malware Detection Framework Based on Intelligent Classification Models

As mobile devices are increasingly used, the number of users employing the Android mobile platform has been on the rise. As a result, malware detection on the Android mobile platform is on the increase. There are various malware detection tools; however, the growing variety of malware represents a major threat to conventional malware detection techniques. In this paper, we propose a stronger framework for malware detection on the Android mobile platform by incorporating intelligent classification techniques. Our framework comprises incorporating the XGBoost Classifier, a classifier that excels in dealing with vast data sets and has a low possibility of overfitting, along with other popular classifiers, including Naive Bayes, K-Nearest Neighbor (KNN), Decision Tree, and Logistic Regression. The system relies on the use of static features that it collects from the Android application package, which include the request for permissions and the calls made to the API. The classification of the application is based on whether it is harmless or malicious. The results clearly show that the XGBoost Classifier obtains an accuracy level of 100%, making it the most outstanding in terms of precision, recall, and the F1-score, offering the new standard in the classification of Android malware. The new framework guarantees the reliability and scalability of Android devices against malicious applications. The limitations and challenges facing the existing methods and proposals to improve the new trend in Android malware classification have also been discussed.