Towards Permission Request Prediction on Mobile Apps via Structure Feature Learning.

摘要: The popularity of mobile apps has posed severe privacy risks to users because many permissions are over-claimed. In this work, we explore the techniques that can automatically predict permission requests a new app based on its functionality and textual description information, which help be aware apps. Our framework formalizes prediction problem as multi-label learning problem, where regularized structure feature is utilized capture relations among descriptions, permissions, category. result learned using our approach. We evaluate approach 173 from 11,067 across 30 categories. Extensive experiment results indicate method consistently provides better performance (3%-5% improvement in terms F1 score), when compared other state-of-the-art methods.