Characterizing User Platforms for Video Streaming in Broadband Networks

TL;DR

This work develops a methodology to identify user platforms for video streams from four popular providers by analyzing network traffic in real-time by developing a classification pipeline that uses 62 attributes extracted from the handshake messages to determine the user device and software agent of video flows with over 96% accuracy.

Abstract

Internet Service Providers (ISPs) bear the brunt of being the first port of call for poor video streaming experience. ISPs can benefit from knowing the user's device type (e.g., Android, iOS) and software agent (e.g., native app, Chrome) to troubleshoot platform-specific issues, plan capacity and create custom bundles. Unfortunately, encryption and NAT have limited ISPs' visibility into user platforms across video streaming providers. We develop a methodology to identify user platforms for video streams from four popular providers, namely YouTube, Netflix, Disney, and Amazon, by analyzing network traffic in real-time. First, we study the anatomy of the connection establishment process to show how TCP/QUIC and TLS handshakes vary across user platforms. We then develop a classification pipeline that uses 62 attributes extracted from the handshake messages to determine the user device and software agent of video flows with over 96% accuracy. Our method is evaluated and deployed in a large campus network (mimicking a residential broadband network) serving users including dormitory residents. Analysis of 100+ million video streams over a four-month period reveals insights into the mix of user platforms across the video providers, variations in bandwidth consumption across operating systems and browsers, and differences in peak hours of usage.

Figures (14)

• Figure 1: Experimental setup for video streaming traffic trace collection.
• Figure 2: Anatomy of network communication for a video streaming session.
• Figure 3: Number of unique values (left blue in log scale) and number of user platforms with different value distributions (right purple in linear scale) for each handshake field in YouTube flows over QUIC.
• Figure 4: Packet processing pipeline for classification of video streaming user platforms.
• Figure 5: Attribute importance in different classification objectives including user platforms, only device types, or only software agents for YouTube (a) QUIC and (b) TCP flows. The level of preprocessing required (annotated by low-, medium- or high-cost) and classification objectives are represented by different colors and patterns, respectively.