11
15. 5. 2016.
Classifying anomalous events in BGP datasets
Border Gateway Protocol (BGP), which enables Internet interconnectivity, is susceptible to various anomalous events that may affect the Internet performance. Understanding the nature of anomalous events (unintentional or malicious) and their effects helps classify future events and improve the Internet robustness. Determining the rate and causes of these anomalous events is important for assessing loss of data and connectivity. BGP update messages contain network reachability information stored in a Routing Information Base (RIB). In this paper, we use datasets of known malicious attacks and a power outage event and employ machine learning algorithms to identify traffic anomalies.