: Using client-specific QUIC headers to identify operating systems or specific device types.
: The original full dataset spans one month , contains over 153 million flows , and was derived from 27 TB of raw traffic.
: The full dataset identifies 102 service classes (e.g., streaming, chat, file transfer) and three background classes for out-of-distribution (OOD) detection. KisX_2022_5K.zip
: Contains QUIC user agents and version identifiers, enabling studies on large-scale QUIC deployment and device recognition. Primary Research Applications
The dataset is a critical resource for researchers studying modern web protocols, encrypted traffic analysis, and machine learning-based network monitoring. : Using client-specific QUIC headers to identify operating
: Collected from 100 Gbps backbone lines of a large ISP connecting roughly 500 institutions and half a million users.
: Designing and evaluating algorithms to identify specific web services (like Netflix or WhatsApp) even when encrypted. : Contains QUIC user agents and version identifiers,
: Training models to detect malicious traffic or "unknown" services that differ from baseline behavior.