Analyzing virus propagation in complex networks.


Today’s networks are very complex and it has become difficult to predict computer attacks. IoT networks are built up from a very large number of devices using a variety of applications, that causes their attack surface to continuously increase and puts them at a high security risk. Therefore, limiting damage spread against both known and yet unknown attacks becomes a very important concern. Network immunisation is defined as the process of vaccines distribution to nodes in order to stop spreading an infection. However, previous studies, especially in the domain of epidemiology, revealed numerous challenges in computing the necessary metrics to make an accurate decision, due to their computational intractability. Futhermore, the approaches used so far to model infection propagation in complex networks treat all nodes and links uniformly by basing all the metrics or centrality measures on the existence of a link/contact between nodes. But, beyond the existence of a link between two nodes, many other properties of a node, especially user behaviours, play a significant role in the propagation of a virus within a network.

The objective of this master thesis proposal is to use Machine Learning on real datasets to identify and compare different security relevant attributes of a node that could help make accurate decision in mitigating virus propagations, and propose some metrics based on those attributes.

The programming part of this work will focus on using the Python (Networkx, Scikit). The goal will be to test some of the results on generated and real-world graphs in order to evaluate their accuracy.

Kontakt: Farell Folly