Breaking the structure of MaMaDroid

Harel Berger, Amit Dvir, Enrico Mariconti, Chen Hajaj

Abstract: Highlights•Different benign sample ratios affect the detection rate of MaMaDroid.•Evasion attacks on the structure of MaMaDroid are highly successful against it.•Mitigation is achieved using a different ML model and an extended feature set.•New methods to capture the confidence and reliability of a classifier.