作者: Emily Kowalczyk , Atif M. Memon , Myra B. Cohen
DOI: 10.1109/ISSRE.2015.7381837
关键词:
摘要: Recent research in mobile software analysis has begun to combine information extracted from an app's source code and marketplace webpage identify correlated variables validate quality properties such as its intended behavior, trust or suspiciousness. Such work typically involves of one two artifacts the GUI text, user ratings, app description keywords, permission requests, sensitive API calls. However, these studies make assumptions about how various are populated used by developers, which may lead a gap resulting analysis. In this paper, we take step back perform in-depth study 14 popular apps Google Play Store. We have studied set 16 different for each app, conclude that output must be pieced together form complete understanding true behavior. show (1) developers inconsistent where they provide descriptions; (2) artifact alone incomplete information; (3) contain contradictory pieces (4) there is need new analyses, those use image processing; (5) without including analyses advertisement libraries, behavior not defined. addition, number downloads ratings does appear strong predictor overall quality, propagated through versions necessarily indicative current version's