Graphical Inference for Extremal Tail Dependence with Applications to Darknet Traffic - a Capstone Project

Abstract

The work for this project was collaborative and integrated into the ongoing Darknet traffic analysis tools deployed at Merit Network. The Darknet Raw Data, used for analysis, consists of network traffic data captured by Merit's 'ORION' Network Telescope and stored on Google Cloud Platform (GCP). Queries return CSV files, and Darknet captures many modalities of data such as the number of packets, byte size, and number of unique scanners (IP addresses). This information is collected over multiple time intervals including 'Daily', '10 Minutes', and 'Hourly,' and from at least 200 countries in total. The data is then matched to different entities of network traffic, including countries, Autonomous System Numbers (ASNs), and ports. The different types of data are also available as either 'raw' or 'percentile', where 'percentile' is a time series of empirical percentiles (computed over time) of the data for each data modality.