Feature selection and ensemble of classifiers for Android malware detection

In recent years the Android Operating System (OS) has become one of the major stakeholders in the smartphone market. The growing consumers' adoption of Android has also brought many security concerns as the number of malicious applications targeting this OS has dramatically increased. Current malware detection methods include static and dynamic analysis. In this work, a set of results obtained for malware classification through machine learning techniques are presented. Although, the presented approach analyzes data obtained through static analysis techniques as other approaches, it differs from previous works by presenting detailed descriptions of the data sets characteristics, feature extraction and selection processes, the size of the training sample set, cross validation and validation sets are specified, providing explicit evidence for classification improvement. Even more, a comparative analysis of various ensembles is presented, having as objective to determine the best combination of classifiers based on the evaluation of the classification results.