Researchers, policymakers and law enforcement agencies across the globe struggle to find effective strategies to control criminal networks. The effectiveness of disruption strategies is known to depend on both network topology and network resilience. However, as these criminal networks operate in secrecy, data-driven knowledge concerning the effectiveness of different criminal network disruption strategies is very limited. By combining computational modeling and social network analysis with unique criminal network intelligence data from the Dutch Police, we discovered, in contrast to common belief, that criminal networks might even become ‘stronger’, after targeted attacks. On the other hand increased efficiency within criminal networks decreases its internal security, thus offering opportunities for law enforcement agencies to target these networks more deliberately. Our results emphasize the importance of criminal network interventions at an early stage, before the network gets a chance to (re-)organize to maximum resilience. In the end disruption strategies force criminal networks to become more exposed, which causes successful network disruption to become a long-term effort.

Taking precautions before or during the start of a virus outbreak can heavily reduce the number of infected. The question which individuals should be immunized in order to mitigate the impact of the virus on the rest of population has received quite some attention in the literature. The dynamics of the of a virus spread through a population is often represented as information spread over a complex network. The strategies commonly proposed to determine which nodes are to be selected for immunization often involve only one centrality measure at a time, while often the topology of the network seems to suggest that a single metric is insufficient to capture the influence of a node entirely.In this work we present a generic method based on a genetic algorithm (GA) which does not rely explicitly on any centrality measures during its search but only exploits this type of information to narrow the search space. The fitness of an individual is defined as the estimated expected number of infections of a virus following SIR dynamics. The proposed method is evaluated on two contact networks: the Goodreau's Faux Mesa high school and the US air transportation network. The GA method manages to outperform the most common strategies based on a single metric for the air transportation network and its performance is comparable with the best performing strategy for the high school network.

No abstract

No abstract

No abstract