Security Semantics Modeling with Progressive Distillation

Petridish is proposed which generates discriminative models against the repacked malicious apps and these representative models of malicious semantics can be progressively distilled with malign and benign samples and can further detect repacks malicious apps.