Artificial Intelligence

Spreading failures and how to fight it

Wednesday, November 15, 2017, 11:00am - 12:00pm PSTiCal
6th floor large conference room
This event is open to the public.
AI Seminar - Interview talk
Goran Muric

Networks are used for modeling numerous technical, social or biological systems. Regardless of the network model and the purpose of the particular network many of them are prone to various types of spreading processes either desired or otherwise. Spreading is observed in social networks in form of the rumors, news or even epidemics among humans. In the case of the telecommunication networks, we observe the data broadcast or failures propagation due to the node breakdown. The latter could manifest in the service outage on a large scale.

In this talk, the focus will be given to two types of spreading failures: Epidemics and Cascades. Those failures share some similarities as they both originate usually in a small fraction of nodes and they both spread through the network often causing a global collapse. However, the cause, the nature and therefore the modelling of those failures differ.

First, the epidemic phenomena will be discussed. I will present the measure called Node Imposed Response (NiR), which accurately evaluates node spreading power. In my research, I was able to utilize the system-theoretic approach considering the network as a Linear Time- Invariant system. By observing the system response we can quantify the importance of each node.

Then, cascading failures will be discussed. Any system which can be modeled using an interdependence graph with limited capacity of either nodes or edges is prone to cascades. Here, I focus on a case study of a particular backbone communication network, European National Research and Education Networks (NREN). Numerical simulations are used to model the cascading failures within the network and to identify the weak points and critical areas. Some protection strategies will be presented as well.

« Return to Events