Whenever possible, we share not only our code, but also the datasets with the scientific community, that this page points to.
AppClassNet is a carrier-grade dataset for traffic classification and application identification research, containing millions of labeled samples from hundreds of applications -- the networking equivalent of the ImageNet dataset!
Maps of IP-ID behaviors that are prevalent in the Internet. Censuses at IP/24 level, along with training set with manual ground truth.
A collection of award winning datasets including both automated collection, as well as large-scale real users campaign.
Over 20GB of controlled experiments with different congestion control algorithms, showing an interplay of data plane throughput vs control plane delay in the performance of distributed applications.
Maps of anycast IPv4 enumeration and geolocalization. Monthly censuses at IP/24 level, lists, Google maps and more!
Methodology to infer queuing delay from simple non-intrusive measurement, along with Internet-wide measurement campaign for BitTorrent hosts and other targets