PERMINAPI: ANDROID MALWARE DETECTION USING PERMISSIONS, INTENTS AND API CALLS

Abstract

Android malware poses a significant threat to mobile devices, necessitating the development of robust detection systems. In this study, we propose a novel approach to Android malware detection leveraging Permission API calls and intents as features for analysis. Our research demonstrates the effectiveness of ensemble learning techniques, including AdaBoost, Gradient Boosting, and Histogram Gradient Boosting, along with a Voting Classifier, in improving detection accuracy, achieving an overall accuracy of 98.7%. We employ k-fold analysis and model training using machine learning algorithms such as Decision Tree, KNN, SVM, and Naïve Bayes to ensure the reliability and robustness of our detection system. By combining the outputs of diverse classifiers through ensemble learning, we harness the collective strengths of individual models, leading to superior detection performance. While our findings show promising results, future research avenues include exploring dynamic feature extraction and integrating deep learning approaches to further enhance the resilience and efficacy of Android malware detection systems. Overall, our study contributes to the advancement of Android security measures, ensuring mobile devices are better protected against evolving threats in the digital landscape.