• 목록
• 답변
• 글쓰기
• 게시판 리스트 옵션
  • 수정
  • 삭제

Within the face of market uncertainty, illicit actors on the darkweb mitigate threat by displacing their operations throughout digital marketplaces. In this examine, we reconstruct market networks created by vendor displacement to examine how digital marketplaces are connected on the darkweb and determine the properties that drive vendor flows before and after a legislation enforcement disruption. Findings present that vendors’ motion throughout digital marketplaces creates a extremely related ecosystem; nearly all markets are directly or indirectly linked. These network characteristics remain stable following a law enforcement operation; prior vendor flows predict vendor motion before and after the interdiction. The findings inform work on collective patterns in offender choice-making and lengthen discussions of displacement into digital spaces.

INTRODUCTION

The emergence of digital marketplaces for the sale of illicit goods has transformed the illicit financial system. Digital marketplaces provide centralized platforms for sellers to advertise their merchandise, join with consumers, and increase their clientele. These marketplaces enable new and completely virtual transactions and complement illicit exchanges that happen offline (Leukfeldt et al. 2017).

Digital marketplaces will not be a brand new phenomenon, but evidence reveals that solely recently have distributors begun to displace their operations across multiple marketplaces at greater rates (Ladegaard 2020). The motion of distributors across digital marketplaces suggests they have develop into more and more interdependent; that's, what happens in one market affects the marketplaces around it. Law enforcement interventions, including the seizure of a market, influence surrounding markets, displacing vendors to different platforms. The move of ‘market refugees’ from seized to neighbouring markets has been identified as one of many focal mechanisms by means of which the online economic system has remained resilient to interventions (Ladegaard 2020). Vendors can maintain their on-line identities and reconnect with current and new shoppers on similarly situated digital platforms.

Crime displacement is central to criminological scholarship. Where offenders resume their illicit activities following an intervention sheds perception into the emergence of scorching spots and the ability to deter crime (Braga et al. 2019). Yet, we know little about what motivates offenders’ decisions to maneuver their illicit actions to a brand new location-physical or in any other case. Digital marketplaces offer a novel opportunity to increase discussions of crime displacement to on-line environments. Vendors, their merchandise, and transactions typically depart a document, providing mass digital traces across illicit marketplaces and enormous populations of vendors as they unfold. Digital records from on-line marketplaces supply a novel opportunity to analyze crime displacement, allowing us to pinpoint the place crime moves to and the pathways it takes to get there. This extends current discussions of displacement and offender resolution-making to incorporate the place offenders transfer to (additionally see Hatten and Piza 2021).

The current examine adopts a community strategy to raised perceive how digital marketplaces are connected through vendor displacement and assess vendors’ choices to maneuver between markets. Specifically, we ask two interrelated questions: 1) how are digital marketplaces on the darkweb related by way of vendor flows, and 2) does the overarching structure of the network help explain vendor flows earlier than and after a legislation enforcement intervention? To reply these questions, we reconstruct vendor flows across digital marketplaces on the darkweb and look at the connectivity of those marketplaces before and after a major interdiction. We then use exponential random graph fashions to determine the correlates of vendor flows and assess whether the drivers of vendor motion are disrupted following a law enforcement intervention. Together, the study aims to inform broader processes about crime displacement as it extends to digital areas.

We start with a evaluation of digital marketplaces on the darkweb with a concentrate on their maturation from more centralized to decentralized illicit economies. We then join this work with research on the impression of interdictions on darknet markets, theoretically grounding our discussion in rational selection and social studying theories. We then detail a mass longitudinal knowledge collection effort to trace vendor flows across multiple giant-scale marketplaces and the social network strategies used to examine the connectivity of this darknet ecosystem. After looking on the aggregate patterns driving vendor flows, we evaluate the affect of a legislation enforcement seizure on vendor motion. We conclude by discussing the implications of the findings for advancing criminological theory on crime displacement and offender determination-making.

CRIME DISPLACEMENT IN DIGITAL Spaces

Digital marketplaces on the darkweb

In 2011, Silk Road turned one of the primary giant-scale marketplaces to promote illicit goods on the darkweb. Adopting an identical infrastructure to legal e-commerce sites, corresponding to Amazon and eBay, it set the stage for the trade of illicit goods, facilitating greater than $300k in transactions day by day (Barratt 2012; Soska and Christin 2015). At its launch, Silk Road was one in all a handful of marketplaces providing an online platform for illicit e-commerce; nevertheless, its success was accompanied by the emergence of rivals and its downfall even more so. In the months following the marketplace’s seizure, a number of different marketplaces emerged to fill its void (Soska and Christin 2015), a pattern that has since continued (Van Buskirk et al. 2017).

Although digital marketplaces on the darknet are extremely unstable, hardly ever surviving greater than a 12 months (Branwen 2019), the larger darknet economic system is resilient to external shocks. Much of the scholarship on the impact of law enforcement disruptions have found the stock of illicit transactions, the quantity of distributors, and the variety of markets recovers comparatively shortly after market seizures. As an example, Décary-Hétu and Giommoni (2017) observed that a big-scale seizure led to preliminary sharp drops in the number of transactions and new vendors registering on e-commerce websites; nonetheless, had been restored to comparable levels within a number of months of the intervention (also see Van Buskirk et al. 2017). Likewise, Ladegaard (2019) found that while a regulation enforcement crackdown led to a significant reduction within the variety of out there markets, the stock of markets returned to the identical level 6 months following the operation and elevated a 12 months and a half later.

Indeed, somewhat than cripple the darknet economy, recent studies suggest that shocks to digital marketplaces have elevated their interdependency. Markets have turn out to be increasingly interdependent because vendors are more likely to cross-record their products across multiple marketplaces. One Europol official, commenting on this phenomenon, observed that ‘[distributors] don’t just function on one market, they cover the full spectrum of the darkish web’ (Barrett 2020). Per this commentary, scholars have documented giant numbers of distributors promoting their merchandise across a number of marketplaces (Décary-Hétu and Giommoni 2017; Ladegaard 2019; 2020; Norbutas et al. 2020).

In some of the persuasive accounts of the affect of law enforcement interventions on vendor displacement, Ladegaard (2020) documented the widespread adoption of authentication techniques across digital marketplaces after a serious disruption. Authentication methods allowed marketplaces to validate vendors’ on-line identities, growing the convenience of shifting between markets and bringing their online reputations with them. Analysing vendor migration throughout three markets, Ladegaard (2020) discovered that many newly registered distributors had migrated from lately seized digital marketplaces. In effect, the intervention triggered marketplaces’ adoption of authentication methods, increasing the flexibility of illicit actors to navigate between what had been as soon as unbiased marketplaces. In addition, the intervention additionally led to an uptick in the number of accessible directories or ‘information hubs’ that provide lists of energetic markets, additional growing the sources from which vendors could draw on to make knowledgeable selections on where to arrange store. These adaptations enabled illicit marketplaces to resemble legal ones more closely. Online identities might be verified, and customers might seek the advice of directories with up-to-date listings of energetic markets.1

Crime displacement, rational selection and offender networks

Crime displacement, which incorporates where people resume their actions after an intervention, is of central theoretical significance to scholarship on crime and criminal justice. Prior analysis exhibits that crime discount efforts often lead to displacement (Reppetto 1976; Gabor 1981), with spatial relocation the most common response (Rossmo and Summers 2021). Where offenders transfer to is theoretically informed by rational selection theory and to a sure extent, social studying concept.

Rational alternative idea views offenders as choice-makers who engage in a price-benefit analysis of the anticipated dangers and rewards of engaging in a criminal act, together with the decision of the place to offend (Becker 1968; Clarke and Felson 1993). In applying rational choice idea to the examine of illicit markets, Reuter and Kleiman (1986) highlight the salient position of perceived rewards and prices related to illicit market exercise, including earnings, incapacitation, and lack of product. Indeed, this same financial calculus has been discovered to underlie the choice-making of actors on digital platforms, including the choice to transition to online markets from offline markets, where earnings are seen as higher and dangers as decrease (Décary-Hétu and Giommoni 2017; Martin et al. 2020).

More not too long ago, scholars have emphasized that offender decision-making does not happen in a vacuum but is knowledgeable by the behaviours and actions of others. In criminology, previous work has found that peers shape the anticipated dangers associated with engaging in crime (Stafford and Warr 1993; Pogarsky et al. 2004; McGloin and Thomas, 2016), perceived advantages (Warr 2002) as well as the talents and alternatives to commit crimes (Weerman 2003; Morselli et al. 2006). The role of peers in shaping offender selections is a core tenet of social learning theory, which emphasizes that people model the behaviours of those around them. Indeed, social studying idea highlights friends as a key reference group from which individuals observe and be taught criminal and delinquent behaviours (Bandura 1978; Akers 2011). Consistent with social learning theories, community frameworks supply an essential device to understand the position of peers on behaviours, with its place to begin the premise that individuals’ actions and beliefs rely upon the actions of others in their networks (Wasserman and Faust 1994).

In illicit on-line markets, the function of offenders’ networks is obvious. Online communities present individuals with access to a pool of peers who inform individuals’ risk of participating in illicit activity (Holt et al. 2008; Aldridge and Askew 2017). Past work has provided anecdotal proof that vendor selections to move to new marketplaces are made collectively (Ladegaard 2020). Moeller et al. (2017) succinctly summarized this phenomenon with a quote from a darknet information forum, ‘If Silk Road is down, everybody moves to Agora, if Agora is down everybody strikes to Evo … and so on […] the DNM’s person base could be very herd like’ (p. 1,434). Together, these works recommend that offenders weigh the costs and advantages of illicit activity and depend on their peer networks for informing their resolution calculus, including where to promote their illicit products.

Although prior work suggests offenders’ draw from their peers to pick out illicit marketplaces, there's a notable gap in empirical work investigating exactly how friends shape vendor flows across markets. This work suggests that peers function essential behavioural models, providing sources of information to evaluate a market’s advantages and prices. Instances where vendors see many of their friends on a market can increase the anticipated benefits (e.g., seeing that different vendors have selected the platform as a helpful place to conduct their business) and scale back perceived prices (e.g., signalling trust in the site as not a rip-off and offering a public display that they have not been arrested) (Ladegaard 2020, p. 13). Alternatively, where people see few of their friends in the marketplace may improve a site’s perceived risk and dependability. For instance, marketplaces with few of their peers might cue a site that has been planted by agents seeking to observe vendor behaviours or indicating there are few consumers on these websites.

In sum, drawing from past theoretical work that contends peers serve as necessary behavioural models from which to observe and learn offending behaviours, and more moderen work that finds illicit market members draw from their friends to evaluate the costs and benefits of illicit activities, we expect vendors’ friends to play an necessary position in shaping online behaviours. Specifically, we count on vendors to maneuver to marketplaces where their peers have moved to previously, leading to the next hypothesis:

Hypothesis 1: Vendor flows usually tend to occur between marketplaces where vendors’ friends have moved to prior to now.

Further, drawing on previous research that emphasizes disruptions increase vendor motion across marketplaces, we anticipate this relationship to strengthen following a regulation enforcement intervention. We'd anticipate the anticipated costs of taking part in illicit activity to be heightened with elevated attention from legislation enforcement. In these contexts, vendors could also be extra threat-averse and more more likely to depend on their peer community to establish trusted sites, following these vendors who weren't detected up to now shutdown. Indeed, prior work has shown that status and belief take on a higher market value after a disruption (Duxbury and Haynie 2020). This line of labor led to the following speculation:

Hypothesis 2: A regulation enforcement disruption will strengthen the relationship between current vendor flows and the place vendors peers’ have moved to prior to now.

Examining VENDOR FLOWS BETWEEN DIGITAL MARKETPLACES

The current research empirically checks these hypotheses by reconstructing vendor flows across digital marketplaces earlier than and after a major law enforcement interdiction. Prior research on crime displacement has primarily focused on whether or not interventions reduce crime or relocate it to other areas (Hatten and Piza 2021). Here, we look at a big sample of offenders and explore the properties that lead them to maneuver to particular on-line areas. In doing so, we seek to maneuver the scholarship on crime displacement forward, substantively and methodologically, by looking at how vendor movement connects digital marketplaces and assessing how the construction of market networks shapes collective patterns in offender choice-making.

Theoretically, our study draws from rational selection and social studying theories to raised understand offender resolution-making and crime displacement in online areas. While early scholars emphasized the necessity to study where (and when) crime occurs (Felson 2006), an absence of detailed knowledge precluded these efforts. The digital landscape affords a new supply of knowledge to investigate offenders’ alternative structures and provide insight into the fundamental determinants of offender displacement patterns. In the present study, we explicitly take a look at whether or not vendors’ choices on the place to sell their products is modelled off the behaviour of their peers. Our outcomes shed gentle on the processes via which vendors transfer to different illicit marketplaces, with a give attention to the financial and social forces that structure these selections.

Methodologically, a community approach allows us to explore questions central to scholarship on crime displacement. The questions being raised on online platforms usually are not new. Crime displacement has been studied for decades, with a lot of this literature specializing in the impression of crime discount efforts on the movement of crime to new areas (Weisburd et al. 2006; Braga et al. 2019), and more recent functions on where offenders transfer to (Hatten and Piza 2021). Specifically, we conceive of marketplaces as a network by which individual e-commerce sites are nodes, and the motion of sellers between websites are edges. We then use exponential random graph models to look at the drivers of vendor motion earlier than and after a law enforcement seizure of considered one of the most important markets. In doing so, we show how a network strategy supplies a novel lens by which to discover the etiology of crime displacement.

DIGITAL Trace Data ON THE DARKWEB

The information for this paper comes from English-language marketplaces that promote stolen information products hosted on the darkweb. Stolen information products are defined right here as fraudulent documents (e.g., drivers’ licenses, passports), monetary objects (e.g., financial institution accounts, credit score cards), counterfeit currencies, companies to steal data (e.g., account crackers, injectors), and tutorials or guides associated to any of the previous categories. Because some markets don't classify product listings or misclassify listings, we used a set of keywords to extract the relevant listings for the analysis (see Appendix I for a full record of key phrases). The data only includes marketplaces with more than one vendor and more than a hundred stolen data listings.

Marketplaces meeting these criteria had been identified by consulting market directories, web sites that listing lively markets on the darkweb and the onion.links to entry them. These web sites provide a beneficial useful resource for vendors and buyers to determine up-to-date information on markets, together with their links, as markets could switch their onion.hyperlink in efforts to elude legislation enforcement or other hostile actors. As well as, marketplaces have been situated by consulting standard forums on the darkweb for discussions of recent markets. Digital records from every marketplace had been then compiled into a structured database utilizing net-scraping and parsing instruments that extracted all publicly accessible product listings, and vendor profiles pertaining to stolen knowledge gadgets (Wu et al. 2019). Our remaining sample comprises 17 markets, 979 unique vendor aliases, and 221,094 product listings over an roughly 12-week period from 15 November 2020, to 9 February 2021.

Methodological barriers largely explain why prior research on the networks of digital markets is proscribed. To assess vendor flows requires capturing vendor exercise across a big sample of digital marketplaces, demanding information throughout multiple platforms, every with hundreds of information factors with completely different infrastructure that can change over time. Because few comprehensive longitudinal datasets across a number of markets exist, these analyses have yet to be carried out. However, amassing knowledge from multiple markets creates empirical obstacles, and the bounds to our approach ought to be noted.

First, marketplaces on the darkweb are notoriously unstable. Markets typically go down for upkeep and will not be accessible for prolonged durations. Due to this, we knowingly omit some listings if the market went down during the scraping interval. Our data assortment method partially overcomes this limitation, as we scraped the markets weekly after which aggregated this data over four weeks, offering more complete data factors. However, we could also be missing listings that went up after which had been taken down within shorter time intervals. Relatedly, we also confronted issues with our own scrapers with the seized market, DarkMarket, not fully scraped within the three weeks prior to it being shut down.

One different limitation that could doubtlessly affect our evaluation should be noted. Our information only accommodates information on vendors’ on-line aliases. It is possible that distributors use completely different aliases across marketplaces or that aliases are ‘mimicked’ by others in efforts to rip-off patrons, and there is some proof of this impact (van Wegberg and Verburgh 2018; Martin et al. 2020). However, latest work means that the adoption of vendor verification processes by web site directors has restricted this chance (Ladegaard 2020; Norbutas et al. 2020), and others have proven that vendor aliases serve as a valid proxy for figuring out vendors’ distinctive identities (Broséus et al. 2016; van Wegberg and Verburgh 2018). Indeed, vendors’ aliases present ‘brand recognition’, and are straight tied to their online reputations, considered one of the primary methods prospects select sellers (Duxbury and Haynie 2018). Although not excellent, within the absence of extra dependable approaches we observe previous work (Décary-Hétu and Giommoni 2017; Ladegaard 2018; 2019) and deal with each vendor alias as distinctive. In doing so, we are conservative in our method, requiring precise matches of vendor aliases to be labeled as the same vendor.

To assist interpret our quantitative findings, we additionally reached out to vendors to conduct interviews on the elements that structured their choices to set up storefronts on digital marketplaces. We recruited vendors who made not less than one sale on a darknet market in the month previous the recruitment message. In whole, 865 unique distributors fitting these criteria have been recognized. As a result of market volatility, our research group was only capable of contact 360 distributors across 12 markets between 4 April 2021 and 1 May 2021 and asked to participate in an asynchronous interview on an encrypted platform of their alternative. Follow-up messages were despatched two weeks after the primary participation request. From the 360 distributors contacted, twelve replied. Of those twelve, one completed the complete interview, and one accomplished a partial interview. Content from these interviews is included to provide perception into the decision-making processes underpinning vendor motion; however, we emphasize our restricted sample, which we return to in the restrictions.

ANALYTIC Approach

Our evaluation focuses on the social networks created by vendor flows wherein the nodes symbolize markets, and the ties characterize the stock of distributors who transfer between any set of markets. Conceptualizing and measuring vendor flows as market-stage social networks permits us to evaluate the structural features of the community and permit the analyses of the mechanisms driving the structure of the observed market community. We measure the market networks in the 1-month period earlier than and after the seizure of one in every of the largest marketplaces on the darkweb-DarkMarket. We start by describing the structural characteristics of the market networks, together with stability in these constructions over time. This includes properties of the network graph comparable to its overall clustering (density), native clustering (clustering coefficient), and the extent to which vendor motion is centralized round a few key markets (degree centralization). We then use exponential random graph models (ERGMs) to look at the local processes that shape international patterns in the construction of vendor flows, and whether or not these processes change before and after the market seizure.

Seizure of DarkMarket

The seizure of the DarkMarket on eleven January 2021, by Europol authorities intently resembles a long line of enforcement interventions aimed toward curbing illicit exercise on the darkweb. At the time of its operation, DarkMarket was recognized as one of the biggest marketplaces for illicit goods on the darkweb (Europol 2021). Overnight, the location was taken down, with regulation enforcement seizing the servers that hosted the website and arresting the alleged operator of the market. Its takedown gives a singular alternative to test how an intervention impacts vendor flows across markets and is consistent with different research that have examined the affect of legislation enforcement interventions on digital marketplaces (van Buskirk et al. 2017; Décary-Hétu and Giommoni 2017; Ladegaard 2019).

Dependent variable: vendor flows between digital marketplaces

The dependent variable measures the intensity of vendor flows between any two units of digital marketplaces involved in the sale of stolen knowledge. The networks are two-mode network affiliation information that information all markets a vendor marketed stolen data merchandise (vendor-by-market) and the dates they had been recorded as listing these products. The affiliation networks are then transformed into networks of co-affiliation by creating a brand new matrix that information the number of vendors who moved between any pair of markets. The ensuing data is a one-mode network (market-by-market) with the identical market listed within the rows and columns of the matrix. The worth of each cell out there matrix indicates the variety of distributors who handed from the sender market (rows) to the receiver market (columns), allowing us to determine the inventory of distributors who listed stolen data products in a single market (Market A), and then began listing stolen knowledge products on one other (Market B). As such, markets are related if 1) a vendor expanded the number of marketplaces they're on (listed products on Market A at time t and then listed products on Market B at time t + 1), or 2) a vendor left a marketplace and joined a brand new one (discontinued itemizing products on Market A at time t and then began listing products on Market B at time t + 1). Thus, ties between markets are directed and valued, indicating the path of the vendor stream and the depth of the circulation, with more distributors transferring between any two sets of markets having larger values. We measure our dependent variable at two time factors, 1 month before the seizure of DarkMarket (pre-seizure community) and 1 month after the seizure of DarkMarket (publish-seizure community).

To regulate for the truth that sure markets may have higher alternative for greater out-flows based mostly on the entire vendor population on that market, we measure vendor out-flows as the number of vendors who move from the market because the proportion of all distributors available on the market at time t. After calculating the ratio of market out-move to the market vendor inhabitants throughout all pairs of markets, we use quartiles to determine thresholds between markets that send few vendors and those that send many distributors. The quartiles classify the edges into categories primarily based on the intensity of vendor flows, with decrease values indicating a lower proportion of out-flow and higher values indicating a better proportion of out-flow. This method was adopted from analyses of human migration networks to manage for nations of different sizes (Vogtle and Windzio 2016).

Exponential random graph fashions

While the community statistics allow us to describe patterns in vendor flows, the use of ERGMs allows us to test 1) the mechanisms that drive the formation of the market networks and 2) the affect of a legislation enforcement interdiction on disrupting the structure of vendor flows between markets. ERGMs mannequin the probability of tie formation inside the observed network as a function of both actor attributes and traits of the network itself. ERGMs are uniquely suited to reply our research query, as they supply a way to beat the problem of endogeneity that's inherent to community knowledge and thus violates assumptions of conventional regression methods (Robins et al. 2012). ERGMs resolve the issue of non-independence by explicitly modelling how one network tie influences the likelihood of different community ties (Lusher et al. 2013). Further, ERGMs permit us to explicitly take a look at peer results by including network features as covariates in the mannequin. This is essential to the present examine, which aims to straight take a look at whether or not patterns in vendor displacement are influenced by the behaviours of other vendors.

The longitudinal nature of the information provides two analytical approaches for modelling change available in the market networks: 1) a temporal ERGM (TERGM) with binary community data, or 2) two separate ERGMs (pre- and post-seizure) with valued network knowledge and a lagged dyadic covariate for prior network construction. The first option, TERGMs prolong commonplace ERGMs by modelling the extent to which the edges (and non-edges) are stable across observations. However, present purposes of TERGM are restricted to binary knowledge, and thus would probably deal with markets with high and low volumes of vendor out-flows as equivalent, conflating very different marketplace profiles. In distinction, the second possibility, valued ERGMs, extends standard ERGMs by also modelling whether a covariate increases or decreases the value of an edge between network actors (Krivitsky 2012). As such, valued ERGMs allow us to evaluate not only which markets experience vendor flows but in addition the depth of those flows, allowing us to measure the stock of vendor movement across markets.

Valued ERGMs require specifying a reference distribution to mannequin how edge values are distributed among network actors. Here, we use a Poisson-reference distribution to model the general network (Krivitsky 2012). We estimate the probability and intensity of ties forming between markets using two courses of predictors: nodal covariates and structural covariates. Nodal covariates test whether or not actor attributes impression their likelihood of receiving or forming a tie and the intensity of that tie. Nodal covariates are dyad unbiased as the likelihood any pair of nodes could have a network tie will depend on their attributes however is just not conditional on other network ties. Structural covariates test whether properties of the community itself affect the probability any pair of nodes will have a network tie and the depth of that tie. Structural covariates are dyad dependent, with the chance of a tie being modelled as conditional on other network ties. Together, these covariates offer totally different insights into the native processes that dictate collective patterns in vendor flows.

Nodal covariates

Number of vendors is a measure of the number of distinctive vendor aliases on the marketplace at time t. This measure serves as a proxy of market supply and is theoretically knowledgeable by rational selection perspectives, which contend that economic calculations, together with supply and demand, drive illicit activity on and offline (Reuter and Kleiman 1986; Aldridge and Décary-Hétu 2016; Demant et al. 2018; Décary-Hétu and Giommoni 2017). We'd count on greater provide (i.e., extra distributors) to cut back the probability distributors would be a part of an already aggressive market. However, we additionally recognize that the variety of vendors can also impression vendors’ danger evaluation for joining the market, impartial of monetary considerations. Indeed, past work has shown the presence of others impacts the choice to have interaction in illicit exercise, increasing an individual’s perceived anonymity and lowering the anticipated sanctions with participating in the activity (McGloin and Thomas 2016). Thus, is it also attainable to conceive that markets with more vendors will entice further vendors to the market.

Price change is a measure of the extent to which itemizing prices change on the market at time t. We measure the common value change of a product listing by taking the same listing and evaluating its worth at weekly intervals. We measure this across all listings and then take the average over the 4-week interval, offering the typical price change throughout product listings in the marketplace. This measure serves as a proxy of a marketplace’s demand, an method consistent with different studies (Décary-Hétu and Giommoni 2017). Much like our measure of market supply, we draw from the rational choice perspective that reveals distributors are motivated by monetary incentives (Reuter and Kleiman 1986; Martin et al. 2020). We thus expect vendors to be more attracted to marketplaces with increases in demand (price increases) and fewer interested in marketplaces with drops in demand (value decreases).

In addition, directed networks offer the chance to investigate how market covariates impact the chance of sending ties or receiving ties. Thus, for both nodal covariates described-number of vendors and value change-we examine the affect of the nodal attribute on out-degree (the probability a market will send high out-flows of vendors to different markets), and in-degree (the likelihood a market will receive high in-flows of vendors from different markets), allowing us to disentangle vendor decisions to go away old markets, from vendor decisions to affix new ones.

Network covariates

Density is modelled utilizing the sum parameter, which signifies the anticipated worth of a tie between any pair of markets primarily based on the worth of all noticed network ties (Handcock et al. 2021). The sum time period is analogous to an intercept in commonplace regression methods, reflecting the baseline edge value across community actors.

Reciprocity is modelled using the mutual term, which estimates the chance a tie between any pair of network nodes might be reciprocated (Handcock et al. 2021). That is, the extent to which distributors from Market A transfer to Market B additionally influences whether or not distributors from Market B move to Market A. Reciprocity is a well-established community process that may impression community structure, serving as an important control for estimating structural processes.

Transitivity is measured using the transitiveweights term, which examines whether or not a tie worth in the community may very well be defined by triad closure. Transitivity occurs in networks when ties between two sets of actors improve the likelihood of a tie between a 3rd actor. In our case, transitivity permits us to check whether or not vendor flows are doubtless to maneuver between markets that have a tie in widespread, and thus whether or not clustering dictates how vendors’ transfer between markets. Prior research on criminal networks has noticed that illicit networks usually tend to adopt decentralized and secure structures following a law enforcement intervention (Morselli, Giguère and Petit, 2007; Ouellet et al. 2017). However, latest work on digital marketplaces has recommended that distributors are more likely to displace their operations following a market seizure, which would suggest that they develop into extra related and fewer secure. A negative effect for this time period this might support the previous speculation (more secure structures), whereas a optimistic effect would assist the latter hypothesis (more efficient constructions) with greater clustering in network ties.

Prior network construction is our predominant covariate and is modelled utilizing a dyadic covariate term, which entails the adjacency matrix of vendor flows in the preceding four-week period, i.e., a lagged dependent variable. The dyadic covariate term permits us to check the speculation that vendor flows are more likely to happen between markets wherein they've occurred up to now and whether a regulation enforcement operation strengthens or interrupts this peer effect. A constructive and statistically significant effect would recommend that vendor flows are structured by the place their friends moved previously. Should this impact turn out to be stronger within the submit-seizure network, this would recommend vendors enhance their reliance on their friends to determine the place to promote their illicit products.

Results

We current our ends in two stages. The first stage describes the structural options of the digital marketplaces before and after a law enforcement seizure. The primary stage aims to determine the extent to which vendor flows connect the various marketplaces and the options of those networks. The second stage explains the generative processes that led to the noticed networks, presenting the outcomes from the ERGMs. The second stage aims to determine the fundamental explanatory variables associated with vendor displacement across markets and whether or not this changes following a disruption. Across both sections, we complement quantitative findings with accounts from our interviews with vendors.

How vendor flows join digital marketplaces

Figure 1 depicts the market networks before and after a legislation enforcement seizure. Each node within the network represents a market involved in the sale of stolen knowledge on the darknet. The dimensions of the node signifies the extent to which distributors moved to that market: bigger nodes signal markets that acquired vendor flows from a larger number of markets. The edges present the intensity of the vendor flows between markets, with thicker edges representing a higher stock of distributors shifting between these markets and arrows indicating the course of the flows.

Vendor flows between digital marketplaces on the darknet. Notes: Node dimension indicates a market’s in-diploma. Edge width captures the intensity of vendor flows, with thicker edges indicating the next volume of distributors flowing between any pair of markets and arrows the path of the circulation. One isolate within the pre-seizure community, Yakuza Market, will not be shown.

Figure 1 highlights two key features of the community. First, digital marketplaces on the darkweb are extremely related. The move of distributors throughout digital marketplaces creates a network that hyperlinks nearly all markets into a single part. Nearly all marketplaces are straight or indirectly linked to each other via vendor flows. Second, this connectivity persists before and after a significant law enforcement intervention. Together, this determine gives a first look at the construction of vendor flows throughout digital marketplaces, exhibiting the linked nature of the darknet ecosystem.

Table 1 presents the descriptive statistics for the market networks, offering a more detailed understanding of how vendor flows are distributed across the network. The pre-seizure market network consists of 17 markets and ninety five ties connecting them. A community density of 0.349 before the seizure of DarkMarket indicates that 35 percent of all possible ties between network actors are observed in the market. The clustering coefficient appears at the local connectivity of the market network, the extent to which ties are clustered round actors. A clustering coefficient of 0.676 suggests that there's a relatively excessive diploma of clustering inside markets. Degree centralization signifies whether or not network ties are concentrated around just a few central actors, with increased values indicating larger concentrations (Freeman 1978). In-degree centralization captures the extent to which just a few markets receive the vast majority of ties. In contrast, out-degree centralization captures the extent to which a number of markets send the majority of ties. Previous to the seizure, markets that received distributors tended to be extra centralized with an in-diploma centralization of 0.401. In contrast, markets that despatched vendors tended to be barely more distributed across marketplaces, with an out-diploma centralization of 0.276.

The network construction of vendor flows between digital marketplaces on the darkweb

Consistent with the pre-seizure network, the put up-seizure community consists of 17 markets, but they're higher linked with a higher number of ties between them, 122 edges as compared to 95 edges earlier than the seizure. Although DarkMarket was seized, we include it within the put up-seizure network to observe the out-circulate of vendors to different markets. The submit-seizure market network turns into more linked, with the density growing to 0.449 and the clustering coefficient to 0.838, as compared to the pre-seizure community. This means that vendor flows became more dispersed, with distributors connecting extra of the markets, a finding in step with prior work that means vendor flows increased following an intervention (Ladegaard 2020). While out-diploma centralization will increase slightly across the pre- and put up-seizure period, in-diploma centralization drops barely within the post-seizure period. This suggests markets sending vendors develop into barely more concentrated around a couple of markets, in keeping with the takedown of DarkMarket and huge outflows from this market. In addition, vendor in-flows grow to be barely much less centralized; the community figure confirms this, highlighting a bigger core group of markets that acquired higher vendor in-flows after the legislation enforcement seizure.

The tendency for distributors to move throughout a number of platforms may be seen in a single vendor’s account of how they choose which marketplaces to sell their products: ‘I initially bought grandfathered into one among the top markets locations also known as white home market, thats where all the real players are. From white house i was capable of get vendor bond waived on virtually every different market place’. Another vendor emphasized that having a number of storefronts minimized any concerns about a market going down: ‘i have loads of backup storefronts already energetic and my customers will know the way to find me not tremendous tough.’ This discovering confirms what has been discovered by others, establishing store across multiple marketplaces is facilitated by market directors (waiving vendor charges for established vendors), and is a method for vendors’ coping with the volatility of markets. In the subsequent section, we explore the processes that lead distributors to select particular marketplaces.

The drivers of vendor flows on the darkweb

Table 2 introduces the results for the Poisson ERGMs, which model the intensity of vendor flows between any pair of markets. We estimate two fashions: the predictors of vendor flows pre-seizure (left) and vendor flows post-seizure (right). For both units of fashions, we include the same set of nodal and structural covariates. For the pre- and publish-seizure networks, the prior network construction time period entails the lagged adjacency matrix of vendor flows in the prior four-week period. Within the post-seizure community, this term entails the adjacency matrix of the pre-seizure network. Within the pre-seizure community, this time period entails the adjacency matrix of the market community 4 weeks prior to the pre-seizure community.

Poisson exponential random graph models predicting vendor flows between digital marketplaces

***p < 0.001, **p < 0.01, *p < 0.05, †p < 0.10.

Table 2 reveals that vendor flows previous to the seizure of DarkMarket were guided by the number of vendors and prices. The damaging and significant effect for the number of distributors-sending market indicates that markets with extra vendors had been less prone to experience out-flows of distributors to different marketplaces. The finding that vendors are less probably to move away from markets with a excessive variety of distributors aligns with previous work, which observes individuals’ perceptions of dangers decreases when extra friends are current (McGloin and Thomas 2016). The negative and important effect for value change-sending market signifies that markets that had a drop in listing prices had been more prone to experience out-flows of distributors to other markets. The discovering that markets with a drop in demand are in step with core tenets of rational selection and the properly-established finding that offender resolution-making is structured by monetary motives (Reuter and Kleiman 1986; Martin et al. 2020). Together, these outcomes present that marketplace factors formed vendor decisions to displace their operations but not the place they selected to maneuver to.

By way of the network variables, the reciprocity time period had a unfavourable and vital impact, exhibiting that out-flows of distributors to different markets tended not to be reciprocated from the receiving market. However, the transivity time period had null results on vendor flows, showing no clustering inside the pre-seizure community. In assist of our major hypothesis, the community lag term-prior community construction-had a positive and important impact, indicating that the movement of vendors between markets was guided by the collective patterns of where individuals had moved up to now.

The mannequin of vendor flows after the seizure of DarkMarket, suggests a change in vendor preferences for choosing marketplaces. Specifically, we observe positive and significant effects for each the number of vendors-receiving market and the variety of vendors-sending market, exhibiting that marketplaces with the next variety of distributors have been more likely to experience out-flows and in-flows of distributors. Thus, after a serious marketplace was seized, vendors responded by moving to markets where there have been extra distributors; nonetheless, in addition they left markets that had greater numbers of distributors. The former result's in line with theoretical expectations that distributors would move to sites where there have been extra distributors, doubtlessly signally larger anonymity, where they have been much less more likely to be singled out and hidden inside a larger group. However, the finding that distributors left markets with the next variety of distributors contrasts with what we discovered in the pre-seizure market, potentially suggesting that the disruption may have made vendors extra danger-averse to stay on the identical market, and extra inclined to develop their operations.

Price adjustments remained a major factor for shaping vendor out-flows and helped explain vendor in-flows within the submit-seizure models. After the seizure of DarkMarket, there was a destructive and vital effect for both worth change-receiving market and price change-sending market, indicating vendors have been more seemingly to maneuver to and from markets that had drops in costs. Although counterintuitive at first, this finding could even be partially explained by the tendency for vendors to look for their very own offers, which they will then resell. As an example, one vendor defined, ‘If I see something that’s a very good deal i will buy it just for the sole intention to resell however at all times bulk listings obviously, that’s the way you make money.’ From this perspective, vendors may be interested in marketplaces from which they can also supply their products extra effectively.

Per the pre-seizure model, the reciprocity term is destructive and significant, indicating that vendor flows weren't reciprocated across marketplaces after the intervention. In distinction to the pre-seizure mannequin, the transitivity term is optimistic and vital, indicating that after the seizure there was clustering of vendor flows between markets, with vendors extra probably to move to markets that had a shared market in common. This result is in line with our descriptive findings that confirmed the community became more clustered following the legislation enforcement seizure. Lastly, according to the pre-seizure network, the prior community structure term stays optimistic and vital. This supplies support for our first speculation that distributors have been more doubtless to maneuver to markets that their friends had moved to prior to now. However, we don't find evidence for our second speculation, which expected this relationship to develop into stronger in the post-seizure community. Rather, we discover that vendor flows stayed comparatively stable before and after the intervention.

Discussion

In the present study, we discover that digital marketplaces on the darkweb are highly connected by way of distributors who span multiple platforms. Further, we observe that distributors do not randomly choose into markets, and these micro-preferences produce aggregate level patterns that generate the ecosystem’s construction. Below we element the principle findings of our study and discuss how they build on prior theoretical and empirical work on offender networks and displacement.

The current examine extends investigations of crime displacement and offender decision-making to indicate that the place offenders decide to commit their crimes is formed by their friends. Vendors had been extra doubtless to pick out into marketplaces the place their friends had moved to previously, and this discovering stayed constant earlier than and after a legislation enforcement disruption. This result aligns with larger propositions from social studying idea that emphasize the role of peers in offender resolution-making (Akers 2011). Although our information don't allow us to uncover the mechanisms that underlie peer results, prior analysis provides some clues. Peers form the perceptions of costs and benefits of deviance, together with perceived sanction risk (Stafford and Warr; 1993; Pogarsky et al. 2004; McGloin and Thomas 2016) and the anticipated rewards (Warr 2002). In digital marketplaces, distributors observing their peers transfer to another market might provide cues that the market is reliable. Indeed, students have long emphasised that a dominant driver of illicit market exercise is trust, with patrons more likely to purchase merchandise from trustworthy vendors, more so than the price of the products being purchased (Duxbury and Haynie 2018, also see Diekmann et al. 2014), and reputation takes on a better market value after a disruption (Duxbury and Haynie 2020). Our results counsel that simply as consumers decide up cues on reliable sellers from other buyers’ experiences, distributors also depend on their networks to assess which markets are reliable on which to sell their wares. In essence, seeing their friends transfer to a new marketplace serves as an endorsement of the platform.

As well as, our study’s findings showed that marketplace networks became more related after a law enforcement intervention, a consequence that runs counter to the effectively documented discovering that illicit networks are inclined to adopt more safe and decentralized buildings in the face of threat and uncertainty (Morselli et al. 2007; Ouellet et al. 2017). The completely different responses of criminal networks throughout offline and contexts could also be partially defined by the anonymity afforded by the darknet. A key consideration as to whether or not a community will undertake safe structures hinges on if they've access to trusted individuals or depend on more dangerous affiliates (Morselli et al. 2007). When danger will increase, people could protect themselves by adopting extra secure network positions the place they're less dependent (or connected) to those much less trusted others. In on-line markets, an individual’s id stays hidden to the market participants, and thus their networks are much less topic to concerns that predominate offline criminal exercise. In these anonymous contexts, vendors extra carefully resemble sellers on licit e-commerce sites, counting on online critiques and rankings to establish the standard of their products. When markets turn into extra volatile, distributors can mitigate dangers by already having established a storefront on one other platform where their vendors can simply discover them. Indeed, one in every of our vendor interviews emphasised that establishing multiple storefronts provide ‘backups’, allowing them to mitigate the loss from market closures.

Lastly, we observe that economic calculus drives offenders’ selections on where to promote their merchandise online. Specifically, we found that vendors have been extra seemingly to maneuver to and from marketplaces that lately skilled drops in demand. The discovering that vendors move from marketplaces that skilled drops in demand is in keeping with a rational selection perspective that identifies financial components as weighing heavily in offender resolution-making, with the goal of maximizing earnings (Reuter and Kleiman 1986). However, the discovering that vendors transfer to marketplaces that also expertise drops in demand runs counter to this logic. While counterintuitive at first, this will likely indicate that vendors who have been experiencing a decrease in demand determined to broaden their analysis to other markets, in keeping with prior analysis which has discovered vendors on a number of markets are more likely to reap greater profits (Ladegaard 2020; Norbutas et al. 2020), and vendor interviews expressing how lower costs permit vendors to capitalize by reselling these merchandise on their own terms. Vendors may absorb these costs in the quick-time period, establishing themselves on the platform on the belief that demand will resume later, a proposition in line with past work (Décary-Hétu and Giommoni 2017).

Limitations

Our examine relies on vendors involved within the sale of stolen knowledge merchandise on digital marketplaces on the darkweb. Stolen information gadgets are the second largest category of illicit products on darkweb marketplaces (after drugs); nevertheless, they only signify a subset of all illicit on-line listings (Hutchings and Holt 2015). While we will capture a excessive variety of markets, we do not have knowledge on all vendors energetic on these markets, or all markets active on the darkweb and clearnet. Limiting our analysis to the subset of merchandise on the darkweb gives the mandatory infrastructure to check a number of distributors using the same variables; nevertheless, this could doubtlessly obscure some patterns that may be observed in other settings, and thus findings apply primarily to this context.

Further, our evaluation solely focuses on the affect of a single shock to digital marketplaces on the darkweb-the seizure of DarkMarket on 11 January 2021. However, this only captures one of many legislation enforcement interventions on the darknet. Earlier interventions, including the shutdown of Empire market in August of 2020, should still be creating waves on the darknet where markets and distributors are recovering from these earlier shocks. Relatedly, while darknet marketplaces present troves of information on illicit transactions, they miss data on some of the core covariates of criminality, including offender backgrounds, such as intercourse, and age, which can impression decisions to offend, and where they determine to commit their offences.

Lastly, we emphasize that our interviews depend on a small pattern. Our low response price may be a function of our sampling body, recruitment technique, or a mixture of each. Vendors who sell stolen information products on the darknet might understand the dangers associated with being interviewed as outweighing the rewards. Thus, it is our belief the response price could be improved by growing the rewards (incentivizing members) or lowering the perceived risks (establishing belief and credibility) of participation. In addition, we additionally take be aware of the small samples of current research adopting similar approaches, including the biggest sample of qualitative interviews being thirteen distributors promoting medication on these platforms (Martin et al. 2020). Strategies, equivalent to creating rapport in online areas, together with partnering with established web sites, may partially explain the discrepancies, and we encourage additional work on this area.

CONCLUSION

Our research advances a network framework to understand digital marketplaces as an ecosystem. Drawing from data across multiple marketplaces, we showed illicit marketplaces are highly connected by distributors who move between completely different platforms, and that these networks became more related after a disruption. Investigating the native mechanisms that drove the structure of the market community, we observed that financial concerns together with fluctuations in market demand structured vendor flows between markets. We additionally found that vendor flows have been extra likely to occur between marketplaces the place their peers had moved to up to now. Together, our research demonstrates the importance of economic and social forces, together with peers’ behaviours, to raised understand crime displacement and offender choice-making.

APPENDIX I. Online Stolen Data Key Words

Footnotes

It's important to notice that the rise in ease with which vendors can transfer between digital platforms has resulted in two distinct however related phenomenon: 1) vendors’ cross-use of platforms (situations the place distributors promote their merchandise across multiple marketplaces), and 2) vendors’ migration throughout platforms (instances the place distributors move their product listings from an old market to a new marketplace). While vendors’ cross-use of platforms and migration characterize distinct phenomena, they overlap significantly. Indeed, the volatility of darknet marketplaces has led to increases in distributors operating out of multiple marketplaces and ‘refugees’ who move to new markets as soon as one has shut down. Both phenomena signify movement patterns, the place an offender might transfer to further websites to mitigate risk and broaden their operations, and both phenomena improve the connectivity and dependency between marketplaces. Within the remainder of this article, we use the term vendor motion and circulation to capture situations the place vendors develop their operations or relocate to new markets.

FUNDING

This materials is predicated upon work supported by the U.S. Department of Homeland Security underneath Grant Award Number 17STCIN00001-05-00. The views and conclusions contained in this doc are these of the authors and should not be interpreted as essentially representing the official insurance policies, either expressed or implied, of the U.S. Department of Homeland Security.

REFERENCES

Akers, R. (2011), Social Learning and Social Structure: A General Theory Of Crime And Deviance, Transaction Publishers.

Aldridge, J., and Askew, R. (2017), ‘Delivery Dilemmas: How Drug Cryptomarket Users Identify and Seek to scale back Their Risk of Detection by Law Enforcement’, International Journal of Drug Policy, 41: One hundred and one-109.

Aldridge, J., and Décary-Hétu, D. (2016), ‘Hidden Wholesale: The Drug Diffusing Capacity of Online Drug Cryptomarkets’, International Journal of Drug Policy, 35: 7-15.

Bandura, A. (1978), ‘Social Learning Theory of Aggression’, Journal of Communication, 28(3): 12-29.

Barratt, M. J. (2012), ‘Silk Road: Ebay for Drugs’, Addiction, 107(three): 683-683.

Barrett, B. (2020), ‘179 Arrested in Massive Global Darkweb Takedown’, Wired. Retrieved June 13, 2021. Available at: https://www.wired.com/story/operation-disruptor-179-arrested-international-dark-net-takedown/.

Becker, G. S. (1968), ‘Crime and Punishment: An Economic Approach’, Journal of Political Economy, 76: 169-217.

Braga, A. A., Turchan, B. S., Papachristos, A. V., and Hureau, D. M. (2019), ‘Hot Spots Policing and Crime Reduction: An Update of an Ongoing Systematic Review and Meta-Analysis’, Journal of Experimental Criminology, 15(three): 289-311.

Branwen, G. (2019), ‘Darknet Market Mortality Risks’, Gwern.Net. Retrieved June 13, 2021. Available at: https://www.gwern.net/DNM-survival.

Broséus, J., Rhumorbarbe, D., Mireault, C., Ouellette, V., Crispino, F., and Décary-Hétu, D. (2016), ‘Studying Illicit Drug Trafficking on Darknet Markets: Structure and Organisation from a Canadian Perspective’, Forensic Science International, 264: 7-14.

Clarke, R. V. G., and Felson, M. (1993), Routine Activity and Rational Choice. Transaction Publishers.

Décary-Hétu, D., and Giommoni, L. (2017), ‘Do Police Crackdowns Disrupt Drug Cryptomarkets? A Longitudinal Analysis of the results of Operation Onymous’, Crime, Law and Social Change, 67(1): Fifty five-75.

Demant, J., Munksgaard, R., and Houborg, E. (2018), ‘Personal Use, Social Supply or Redistribution? Cryptomarket Demand on Silk Road 2 and Agora’, Trends in Organized Crime, 21(1): Forty two-sixty one.

Diekmann, A., Jann, B., Przepiorka, W., and Wehrli, S. (2014), ‘Reputation Formation and the Evolution of Cooperation in Anonymous Online Markets’, American Sociological Review, seventy nine(1): Sixty five-85.

Duxbury, S. W., and Haynie, D. L. (2018), ‘Building Them up, Breaking Them Down: Topology, Vendor Selection Patterns, and a Digital Drug Market’s Robustness to Disruption’, Social Networks, 52: 238-250.

Duxbury, S. W., and Haynie, D. L. (2020), ‘The Responsiveness of Criminal Networks to Intentional Attacks: Disrupting Darknet Drug Trade’, PLOS ONE. 15(9): e0238019.

Europol. (2021), ‘DarkMarket: World’s Largest Illegal Darkweb Marketplace Taken Down.’, Europol. Retrieved June 16, 2021. Available at: https://www.europol.europa.eu/newsroom/information/darkmarket-worlds-largest-unlawful-darkish-web-marketplace-taken-down.

Felson, M. (2006). Crime and Nature. SAGE Publications.

Freeman, L. C. (1978), ‘Centrality in Social Networks: Conceptual Clarification’, Social Networks, 1: 215-239.

Gabor, T. (1981), ‘The Crime Displacement Hypothesis: An Empirical Examination’, Crime & Delinquency, 27(three): 390-404.

Handcock, M., Hunter, D., Butts, C., Goodreau, S., Krivitsky, P., and Morris, M. (2021), ergm: Fit, Simulate and Diagnose Exponential-Family Models for Networks. The Statnet Project (https://statnet.org). R package model 4.0.1. Available at: https://CRAN.R-mission.org/package=ergm.

Hatten, D., Piza, E. L. (2021), ‘When Crime Moves Where Does It Go? Analyzing the Spatial Correlates of Robbery Incidents Displaced by a place-Based Policing Intervention’, Journal of Research in Crime and Delinquency, Online First.

Holt, T. J., Blevins, K. R., and Kuhns, J. B. (2008), ‘Examining the Displacement Practices of Johns with On-Line Data’, Journal of Criminal Justice. 36(6): 522-28.

Hutchings, A., and Holt, T. J.. (2015), ‘A Crime Script Analysis of the online Stolen Data Market’, British Journal of Criminology, fifty five(3): 596-614.

Krivitsky, P. N. (2012), ‘Exponential-Family Random Graph Models for Valued Networks’, Electronic Journal of Statistics, 6: 1100-1128.

Ladegaard, I. (2018), ‘We Know Where You're, What You are Doing and We'll Catch You: Testing Deterrence Theory in Digital Drug Markets’, The British Journal of Criminology, 58(2): 414-33.

Ladegaard, I. (2019), ‘Crime Displacement in Digital Drug Markets’, International Journal of Drug Policy, sixty three: 113-21.

Ladegaard, I. (2020), ‘Open Secrecy: How Police Crackdowns and creative Problem-Solving brought Illegal Markets out of the Shadows’, Social Forces, ninety nine: 532-559.

Leukfeldt, R. E., Kleemans, E. R., and Stol, W. P. (2017), ‘Cybercriminal Networks, Social Ties and Online Forums: Social Ties Versus Digital Ties inside Phishing and Malware Networks.’, British Journal of Criminology, 57(three): 704-22.

Lusher, D., Koskinen, J., and Robins, G. (2013), Exponential Random Graph Models for Social Networks: Theory, Methods, and Applications, Cambridge University Press.

Martin, J., Munksgaard, R., Coomber, R., Demant, J., and Barratt, M. J. (2020), ‘Selling Drugs on Darkweb Cryptomarkets: Differentiated Pathways, Risks and Rewards’,