Descriptor: Gaming and Video Streaming Traffic for 5G Dataset (GAViST5G)

This paper introduces a network traffic capture dataset focused on online gaming and video streaming applications, designed to address the challenges in 5G research posed by high throughput requirements and low latency demands. The dataset captures around 200 hours of real-world traffic data (at 3:1 ratio of video streaming to online gaming traffic) from 11 mobile operators across nine countries: Canada, Egypt, Finland, Germany, Kuwait, Norway, Taiwan, the U.A.E., and the U.K; it includes fields such as timestamps, packet sizes, protocols, source and destination IPs, and packet metadata. To improve usability, the dataset was enhanced with geolocation data (country, region, city, longitude, latitude), content provider labels, and round-trip time measurements, offering comprehensive insights into application-specific traffic patterns. The data collection process utilized multiple operating systems, including Windows, Android, and MacOS, ensuring diverse platform coverage. The dataset is stored in CSV format, facilitating easy integration with machine learning workflows for tasks such as quality of service analysis, network slicing optimization, and server selection. In addition to the packet-level data, a second version of the dataset aggregated by the second is provided to support time-series analysis and reduce processing complexity for certain applications. By offering a realistic foundation for model development and evaluation, this dataset serves as a valuable resource for researchers and practitioners seeking to optimize network performance for bandwidth-intensive and latency-sensitive applications.