With the prevalence of smartphones, people nowadays can access a wide variety of services through diverse apps. A good Graphical User Interface (GUI) can make an app more appealing and competitive in app markets. Icon widgets, as an essential part of an app’s GUI, leverage icons to visually convey their functionalities to facilitate user interactions. Whereas, designing intuitive icon widgets can be a non-trivial job. Developers should follow a series of guidelines and make appropriate choices from a plethora of possibilities. Inappropriately designed or misused icons may cause user confusion, lead to wrong operations, and even result in security risks (e.g., revenue loss and privacy leakage). To investigate the problem, we manually checked 9,075 icons of 1,111 top-ranked commercial apps from Google Play and found 640 misleading icons in 312 ( 28%) of these apps. This shows that misleading icons are prevalent among real-world apps, even the top ones. Manually identifying misleading icons to improve app quality is time-consuming and laborious. In this work, we propose the first framework, IconSeer, to automatically detect misleading icons for mobile apps. Our basic idea is to find the discrepancies between the commonly perceived intentions of an icon and the actual functionality of the corresponding icon widget. IconSeer takes an Android app as input and reports potential misleading icons. It is powered by a comprehensive icon-intention mapping constructed by analyzing 268,353 icons collected from 15,571 popular Android apps in Google Play. The mapping includes 179 icon classes and 852 intention classes. Given an icon widget under analysis, IconSeer first employs a pre-trained open-set deep learning model to infer the possible icon class and the potential intentions. IconSeer then extracts developer-specified text properties of the icon widget, which indicate the widget’s actual functionality. Finally, IconSeer determines whether an icon is misleading by comparing the semantic similarity between the inferred intentions and the extracted text properties of the widget. We have evaluated IconSeer on the 1,111 Android apps with manually established ground truth. IconSeer successfully identified 1,172 inconsistencies (with an accuracy of 0.86), among which we further found 482 real misleading icons.

@inproceedings{LiWZ0GW023,
  title = {What You See Is What You Get? It Is Not the Case! Detecting Misleading Icons for Mobile Applications},
  author = {Linlin Li and Ruifeng Wang and Xian Zhan and Ying Wang and Cuiyun Gao and Sinan Wang and Yepang Liu},
  booktitle = {ISSTA},
  year = {2023},
  doi = {10.1145/3597926.3598076}
}