Using Social Network Analysis to disarticulate criminal networks

It has been ascertained that criminal networks, from a structural point of view, trace the biological neural networks that are the most efficient ones in the passage of information from node - neuron - to node. In both cases, the networks are “small world” and “scale free”, which allow them to form short paths even when the networks are large and complex.  It is no coincidence that anti-mafia judge Giovanni Falcone – who was murdered in a bomb attack in 1992 – defined those who, at the top of organized crime, decide its strategies as “very refined minds”.

So how is it possible to distinguish between those who have a leading role, a subordinate one and those who do not belong to a criminal association? Understanding the role played by each node of the network is fundamental because, according to its importance, the police can then decide the type of action to undertake: whether that be a single arrest or a raid. This is the kind of choice that, if wrong, can make it even more difficult to isolate and capture the high profile figures in a criminal organization. What can be done, and what Liotta and his colleagues from the University of Messina and the University of Derby have done, is to overturn the perspective and use another analysis technique: that of social networks.

The Criminal Network of Two Sicilian Clans Reconstructed as a Knowledge Graph
In the first scientific paper, “Disrupting resilient criminal networks through data analysis: The case of Sicilian Mafia,” Liotta and his colleagues reconstructed how the network of relationships of a Mafia clan works. In order to do so, they worked on a real judicial case: they sifted through and manually copied all the data contained in thousands of pages from the acts of the trials against members of the “Mistretta” Mafia family and of the “Batanesi” clan, who had been arrested as part of Operation Montagna in the early 2000s and then sentenced by the Court of Messina. Starting from the two datasets created - and based on wiretaps and stakeouts – they used an algorithm to create a knowledge graph that represented the criminal network of the two clans. Once they had obtained a graphical representation, the topological analysis allowed them to see interactions between the different nodes that would not appear obvious at first glance, and in this way, they could understand the flow of information and interactions between the various affiliates.  The researchers used a metric called Betweeness Centrality that enabled them to evaluate the nodes that play a decisive role in the spreading of information between different segments of the network (the bridge’s nodes). Once these nodes have been identified, targeted attacks can take place in order to weaken the entire network more effectively. This metric does not necessarily measure the shortest path, but the most effective one, which makes it possible to achieve the intended purpose of identifying the person who issues the order to commit a crime, such as murder. Understanding which nodes are fundamental to the graph (those with a higher Betweeness Centrality measure) can help police forces in their decisions concerning who to arrest, and when, and whether to hit several nodes at once by targeted raids.

“By doing so, they can maximize the ability to reduce any communication across the network even if the police do not succeed in arresting the boss because he does not show up in the network except through trusted intermediaries”, Liotta explains, “if you isolate key elements, you give law enforcement time to go after the boss. This way we can minimize the boss’s ability to restore his criminal network.”

The Second Research Path
Liotta and colleagues’ papers have sparked interest in those within the academic sector in the United States who study Mafia criminal networks on American soil. “Some colleagues have asked us for our samples of criminal networks and are doing comparative studies with U.S. datasets”, Liotta says. “We have tried to make a similar request to other organizations in other countries and it has been virtually impossible. Anyway we are willing to broaden our research horizon to different organizational forms.”

To a certain extent, Liotta and colleagues have already done so with a second paper: “Criminal Networks Analysis in Missing Data scenarios through Graph Distances.” In this paper, the authors compared nine different criminal networks: not only the Sicilian Mafia, but also ‘Ndrangheta clans (from the Italian Region of Calabria), drug trafficking groups from Quebec, Stockholm street gangs and terrorists from the Abu Sayyaf group, in the South of the Philippines. Based on data provided by those who had made knowledge graphs of the aforementioned organizations, Liotta and colleagues created algorithms that allow them to generate a “synthetic” criminal network that can be tailored to any organization about which they have little information. “Thanks to these algorithms, we are able to create a kind of approximation, a model of the network we want to dismantle, and then we discover the best operational strategies to unmask key individuals so as to slow down the passage of information between nodes”, he clarifies.

Although the findings from the use of these algorithms allow us to refine the search for hidden nodes in the networks, there is still a lot of work to be done in order to create tools that can be quickly used by police forces all over the world. The effectiveness of these methodologies is directly linked to the availability of operational data, which are typically protected due to the secrecy of the investigation. Prof. Liotta is already working with various research groups on the development of techniques for the anonymization of sensitive data, in order to create more direct collaboration between the university and the special investigation divisions of the police.