A complex networks approach to find latent clusters of terrorist groups

Abstract

Given the extreme heterogeneity of actors and groups participating in terrorist actions, investigating and assessing their characteristics can be important to extract relevant information and enhance the knowledge on their behaviors. The present work will seek to achieve this goal via a complex networks approach. This approach will allow finding latent clusters of similar terror groups using information on their operational characteristics. Specifically, using open access data of terrorist attacks occurred worldwide from 1997 to 2016, we build a multi-partite network that includes terrorist groups and related information on tactics, weapons, targets, active regions. We propose a novel algorithm for cluster formation that expands our earlier work that solely used Gower's coefficient of similarity via the application of Von Neumann entropy for mode-weighting. This novel approach is compared with our previous Gower-based method and a heuristic clustering technique that only focuses on groups' ideologies. The comparative analysis demonstrates that the entropy-based approach tends to reliably reflect the structure of the data that naturally emerges from the baseline Gower-based method. Additionally, it provides interesting results in terms of behavioral and ideological characteristics of terrorist groups. We furthermore show that the ideology-based procedure tends to distort or hide existing patterns. Among the main statistical results, our work reveals that groups belonging to opposite ideologies can share very common behaviors and that Islamist/jihadist groups hold peculiar behavioral characteristics with respect to the others. Limitations and potential work directions are also discussed, introducing the idea of a dynamic entropy-based framework.

Corruption and complexity: a scientific framework for the analysis of corruption networks

“…An additional study that explores 28 years of bill-voting in Brazil shows that the dynamics of co-occurrence networks of similar-voting congressmen reveal patterns that allow for the identification of convicted corrupt politicians and also, for the possibility of predicting or identifying other possible corrupt individuals within the network (Colliri & Zhao, 2019). Noteworthy, the identification of latent criminal groups (Campedelli, Cruickshank & Carley, 2019) and the effective dismantling of their organizational structure (Wandelt, Sun, Feng, Zanin & Havlin, 2018) are relevant and non-trivial subjects in criminal investigations and law enforcement, since empirical evidence has shown that the dismantling process might potentially make these criminal organizations stronger (Duijn, Kashirin & Sloot, 2014). In addition, when it comes to fighting corruption the goal is clear: one not only is looking to describe it post factum, but to predict it (Rumi, Deng & Salim, 2018;Alves, Ribeiro & Rodrigues, 2018;López-Iturriaga & Sanz, 2018;Colonnelli et al, 2019;Colliri & Zhao, 2019).…”

Analysis of Malware Communities Using Multi-Modal Features

“…The methods for collectively clustering the modal graphs are often either diffusion processes (see [17], [19], [20] for examples) or tensor decompositions (see [15], [21]- [24] for examples). Outside of these two main graph-based methods, there are also graph-based methods that use weighted metrics to combine graphs for clustering [25], [26] as well as weight-learning techniques in spectral paradigms to cluster the modal graphs [27] or combining graph decompositions from across modes to then cluster unimodally [16], [28]. While the graph-based methods are often considered the predominant means of multi-modal clustering its not clear how well they can do on more general types of data outside of images or social networks, where the graphs of each mode have similar properties to each other.…”

“…Since modularity is a measure of how much cluster structure is present in a graph, we are selecting graphs that maximize the cluster structure that may be present in the data. We further incorporate the changes to the original algorithm from other works, namely to use asymmetric k-NN graphs for each value of k (wherein an edge exists between two nodes, u and v, if either u ∈ kNN (v) or v ∈ kNN (u)) and the Louvain method for clustering the kNNs [25], [26], [32]. Finally, we note that the Louvain algorithm, while very fast for clustering a graph, is also a greedy, stochastic algorithm that can lead to different clustering outcomes for a graph depending on the initialization of the algorithm.…”

Multiplex networks reveal geographic constraints on illicit wildlife trafficking

“…For instance, spatial signatures of the operation (e.g. where crimes take place) may illuminate how criminal activities are developed, and across which physical and cultural structures they depend upon, among other characteristics (see, Campedelli et al 2019;Graif et al 2014).…”