Collection of datasets with DNS over HTTPS traffic
Kamil Jeřábek, Karel Hynek, Tomáš Čejka, Ondřej Ryšavý
Abstract
Recently, the Internet has adopted the DNS over HTTPS (DoH) resolution mechanism for privacy-aware network applications. As DoH becomes more disseminated, it has also become a network monitoring research topic. For comprehensive evaluation and comparison of developed classifiers, real-world datasets are needed, motivating this contribution. We created a new large-scale collection of datasets consisting of two classes of traffic: i) DoH HTTPS communication and ii) non-DoH HTTPS connections. The DoH traffic is captured for multiple DoH providers and clients to include nuances of various DoH implementations and configurations. The non-DoH HTTPS connections complement the DoH communication aiming to include a wide range of existing network applications. The dataset collection consists of network traffic generated in a controlled environment and traffic captured from a real ISP network. The resulting datasets thus provide real-world network traffic data suitable for evaluating existing classifiers and the development of new methods.