Android malware detection with unbiased confidence guarantees

摘要: Abstract The impressive growth of smartphone devices in combination with the rising ubiquity using mobile platforms for sensitive applications such as Internet banking, have triggered a rapid increase malware. In recent literature, many studies examine Machine Learning techniques, most promising approach malware detection, without however quantifying uncertainty involved their detections. this paper, we address problem by proposing machine learning dynamic analysis that provides provably valid confidence guarantees each detection. Moreover particular hold both malicious and benign classes independently are unaffected any bias data. proposed is based on novel framework, called Conformal Prediction, combined random forests classifier. We its performance large-scale dataset collected installing 1866 4816 real android device. make collection data available to research community. obtained experimental results demonstrate empirical validity, usefulness unbiased nature outputs produced approach.