This dataset was collected from 36 popular Android applications across three distinct network scenarios. The scenarios varied by Internet service provider (ISP), geographic location, device, application version, and individual users. Traffic was generated through human interactions on physical devices for 5–15 minutes using multiple smartphones simultaneously to simulate real traffic from a network supervisor's perspective. The traffic capture method did not require root privileges and filtered out any background traffic. The dataset includes approximately 35 million packets, 399K bidirectional flows, and 29 GB of data. This dataset allows for evaluating model performance under diverse conditions and testing domain adaptation scenarios. The dataset is provided in raw format (PCAP files), to offer researchers flexibility in developing models based on different traffic objects, features, and methods.
Khesal et al. (Sun,) studied this question.