Network visualisation and analysis

• Main Author: Neo, Wei
• Other Authors: Lee Bu Sung, Francis
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science; BGP
• Online Access: https://hdl.handle.net/10356/180849
• Institution: Nanyang Technological University

The Border Gateway Protocol (BGP) plays a vital role in how different networks, knownasAutonomoussystems (AS), communicate across the internet. However, BGP is not perfect and is vulnerable to several types of anomalies like route hijacking, route leakage, and prefix hijacking that can potentially affect both the stability and security of the internet. In this project, the use of Geometric Deep Learning, specifically Graph Neural Net works (GNNs), to detect anomalies in BGP routing data. GNNs, particularly Graph Convolution Networks (GCNs) and GraphAttention Networks (GANs), are well-suited for this task because they can learn from graph-structured data. BGP routing informa tion was represented as a graph-like structure, where the nodes are autonomous systems and the edges are the BGP updates between them to enable the GNN models to capture the complex relationships and patterns in the network. For data extraction BGPStream and BGP Machine Learning (BML) framework was used to extract historical data from many different sources and utilized to process the data in converting it into a graph format that could be fed into the GNN models for training. In summary, the results from the project showed that GNNs particularly GANs out performed GCN in the task of BGP route anomaly detection. The ability of GAN coupled with the attention mechanism to understand the graph-like structure of BGP routing data allowed the accurate identification of unusual events with a higher level of precision compared to the GCN have a strong potential for improving the detection tasks of network anomalies and by taking advantage of the natural structure of BGP routing data, these models could help enhance the security and stability of the internet.