BGP Route Leaks Detection Using Supervised Machine Learning Technique

Abstract

The route leaks problem is considered one of the unsolved Border Gateway Protocol problems for more than fifteen years ago. It has a large negative impact on global internet stability and reliability. This problem is hard to be prevented due to human errors and misconfigurations, and hard to be detected due to the confidentiality of autonomous systems relationships.The paper proposes a new taxonomy to the different types of route leaks depending on their effects on the Border Gateway Protocol traffic, the first real route leaks incidents dataset, and a complete real-time detection system based on a supervised learning classification method. The work compares three classifiers (Decision Tree, Random Forest Trees, and Support Vector Machines). The proposed system prototype can detect and classify route leaks from normal updates with an accuracy of 87% and time complexity of O(NM), where N is the number of prefixes each with M prefix length.