Mobile payment apps have been widely-adopted, which brings great convenience to people's lives. However, at the same time, user's privacy is possibly eavesdropped and maliciously exploited by attackers. In this paper, we consider a possible way for an attacker to monitor people's privacy on a mobile payment app, where the attacker aims to identify the user's financial transactions at the trading stage via analyzing the encrypted network traffic. To achieve this goal, a hierarchical identification system is established, which can acquire users' privacy information in three different manners. First, it identifies the mobile payment app from traffic data, then classifies specific actions on the mobile payment app, and finally, detects the detailed steps within the action. In our proposed system, we extract reliable features from the collected traffic data generated on the mobile payment app, then use a series of well-performing ensemble learning strategies to deal with three identification tasks. Compared with prior works, the experimental results demonstrate that our proposed hierarchical identification system performs better.

The authors declare no conflict of interest.