Communication networks are prone to virus and worms spreading and cascading failures. Recently, a number of social networking worms have spread over public Web sites. Another example is error propagation in routing tables, such as in BGP tables. The immunization and error curing applied to these scenarios are not fast enough. There have been studies on the effect of isolating and curing network elements, however, the proposed strategies are limited to node removals. This paper proposes a link isolation strategy based on the quarantining of susceptible clusters in the network. This strategy aims to maximize the epidemic control while minimizing the impact on the clusters performance. We empirically study the influence of clustering on robustness against epidemics in several real-world and artificial networks. Our results show an average curing rate improvement above 50% for the studied real-world networks under analysis.@en

As communication and network technology has advanced into all areas of human life, large-scale failures of such infrastructures can have a drastic impact on its users, the economy and society in general. Significant attention has been paid to study and improve the dependability of such systems, intending to augment their resilience against challenges. This paper presents a computational approach for a multi-level multi-dimensional study of network robustness, providing hard performance guarantees for networks and helping to assess risk and uncertainty associated with threats. We demonstrate in a case study that such approach is well suited to understood the root causes and subsequent effects of network challenges and therefore obtain deeper insights into the resilience of a network under stress.@en

Several topology generator algorithms have been proposed to match the heavy-tailed behavior of the real Internet AS-level topology. Although evaluation studeis have been presented for each topology generator, each of these studies assumes a different framework. Therefore, it is difficult to compare the differences between the topology generator algorithms. The objective of thsi paper is to fill this gap by comparing representative families of topology generators.@en