Agent-Based Modelling for Criminal Network Interventions

This chapter examines the simulated impact of sequential node deletion on a cryptomarket’s ease of operation. To this extent, computer simulations with incorporated network adaption and preferential selection were leveraged to better understand which strategic intervention(s) were most effective at disrupting the structural integrity of Abraxas’ network structure. Based on the results of the sequential node deletion, random targeting was found to be ineffective across the five outcome measures, producing a minimal and slow disruptive effect. Degree centrality and reputation targeting were the most effective strategies across all five outcome measures, consistently producing near-identical results. The disruption pattern demonstrated by these four targeting strategies was that the proportion of potential measurable values increases or decreases as more actors are removed. These values plateau as the network becomes completely fragmented. Furthermore, it is highly likely that these strategies are interrelated as the specific actors that are targeted are the same or similar. This suggests that, while the stated objectives of these targeting strategies are different, their functional performance is the same or similar. Finally, analyses of the disruptive impact of each targeting strategy show that degree centrality targeting, reputation targeting, total purchase price targeting and unique items bought/sold targeting are each based on a power law in which a small percentage of deleted nodes is responsible for an outsized proportion of the disruptive impact across all five outcome measurements (e.g. 1–5% of deleted nodes were responsible for 45–90% of disruptive impact). This particular study is important for cryptomarket disruption strategies as it demonstrates that the behaviour of an illicit trade network can be modelled (Duxbury & Haynie, 2019) and subsequently vivisected through an evidence-based calculus. Moreover, it provides insight into how law enforcement might approach the curtailment of a cryptomarket. As cryptomarket takedowns and the opportunistic arrest of vendors are not particularly effective in disrupting these entities long term, a carefully calibrated intervention that considers network dynamics, such as preferential selection, is warranted.