Analysis and Evaluation of SafeDroid v2.0, a Framework for Detecting Malicious Android Applications

摘要: Android smartphones have become a vital component of the daily routine millions people, running plethora applications available in official and alternative marketplaces. Although there are many security mechanisms to scan filter malicious applications, malware is still able reach devices end-users. In this paper, we introduce SafeDroid v2.0 framework, that flexible, robust, versatile open-source solution for statically analysing based on machine learning techniques. The main goal our work, besides automated production fully sufficient prediction classification models terms maximum accuracy scores minimum negative errors, offer an out-of-the-box framework can be employed by researchers efficiently experiment find effective solutions: makes it possible test different combinations classifiers, with high degree freedom flexibility choice features consider, such as dataset balance selection. also provides server, generating reports, application, verification produced real-life scenarios. An extensive campaign experiments presented show how competitive results confirm very good performances, even highly unbalanced inputs always limited overhead.