Efficient and Effective Static Android Malware Detection Using Machine Learning

The increasing use of android, its openness and lack of security checks have led to an alarming increase in malware applications. The traditional signature-based detection methods are inefficient against sophisticated malware, and lack scalability resulting in lingering concerns about their reliability. Given the state of affairs, there is an urgent need for a reliable and scalable alternative to signature-based techniques. In this work, we present an effective and reliable machine learning based approach for static android malware detection. We also propose an efficient and effective feature set consisting of 25 features. We achieve an accuracy of 94.68% using Random forest classifier on 20% test size. High recall of 94.67%, precision of 94.68% and f1 score of 94.68% were achieved. We implemented various existing android malware detection schemes and a detailed comparison reveal that the proposed scheme outperforms them all in all classification metrics.