See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/358236970

Seeking anonymity on the Internet: The knowledge accumulation process and

global usage of the Tor network

Article in New Media & Society · January 2022

DOI: 10.1177/14614448211072201

CITATIONS READS

3 136

4 authors:

Zhicong Chen Eric Jardine

Nanjing University Virginia Tech (Virginia Polytechnic Institute and State University)
7 PUBLICATIONS 18 CITATIONS 46 PUBLICATIONS 548 CITATIONS

SEE PROFILE SEE PROFILE

Xiao Fan Liu Jonathan J. H. Zhu

City University of Hong Kong City University of Hong Kong
59 PUBLICATIONS 634 CITATIONS 108 PUBLICATIONS 2,622 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Xiao Fan Liu on 29 April 2023.

The user has requested enhancement of the downloaded file.

1072201
research-article2022
NMS0010.1177/14614448211072201new media & societyChen et al.

Article

new media & society

Seeking anonymity on the

1­–22
© The Author(s) 2022
Article reuse guidelines:
Internet: The knowledge sagepub.com/journals-permissions
DOI: 10.1177/14614448211072201
https://doi.org/10.1177/14614448211072201
accumulation process and journals.sagepub.com/home/nms

global usage of the Tor

network

Zhicong Chen
City University of Hong Kong, Hong Kong

Eric Jardine
Virginia Tech, USA

Xiao Fan Liu

Jonathan J. H. Zhu
City University of Hong Kong, Hong Kong

Abstract
The Onion Router (Tor) network is one of the most prominent technologies for
accessing online resources while preserving anonymity. Effectively employing the
technology is not a trivial process and involves the following steps: (1) motivated by
needs, (2) becoming aware of and learning the technology, and (3) realizing desired
purposes by usage. Using country-level panel data, this study examines the knowledge
accumulation process through which motivated users eventually employ Tor. The results
suggest that Tor is often searched in less free countries for censorship circumvention,
while it is employed for Dark Web activities in more free countries. There is also an
indirect relationship between being aware of the technology and its usage through how-
to knowledge accumulation. This study is the first attempt to understand the role of
knowledge accumulation in the global usage of Tor. The findings provide insights into
the worldwide concerns of online privacy and Dark Web regulation.

Corresponding author:
Xiao Fan Liu, Department of Media and Communication, City University of Hong Kong, 18 Tat Hong
Avenue, Kowloon, Hong Kong SAR, China.
Email: [email protected]
2 new media & society 00(0)

Keywords
Dark Web, innovation-decision process, knowledge accumulation process, online
anonymity, Tor

The Onion Router (Tor) anonymity network, that is, a global overlay network compris-
ing roughly 6000–6500 nodes running an encryption-enhanced routing protocol, is a
specially designed technology for accessing remote computer servers while ensuring that
user identity and location are obfuscated (Dingledine et al., 2004). The Tor network is a
dual-use technology that can be used for both “good” and “bad.” On one hand, it is a
perfect tool for accessing regular web content (“Surface Web”) while overcoming gov-
ernment and corporate surveillance on individuals and promoting Internet freedom. On
the other hand, the Onion/Hidden Services sites (“Dark Web”), which also run regular
web protocols (e.g. HTTPS) but can only be accessed through rendezvous points assigned
internally to the Tor network (Gehl, 2018: 5), are a hotbed for illicit and illegal activities
(Moore and Rid, 2016; Owen and Savage, 2016).
Previous studies examining the Tor network from a social perspective have tended to
focus on three predominant areas: the emergence and legitimacy of subcommunities on
Dark Web platforms (Gehl, 2016; Maddox, 2020); the categorization of available
Onion/Hidden Services contents into legal/illicit categories (Benjamin et al., 2019;
Faizan and Khan, 2019; Moore and Rid, 2016); and the uses and abuses of cryptomar-
kets, as well as their resilience against law enforcement policing efforts (Decary-Hetu
and Giommoni, 2017; Van Buskirk et al., 2017). These areas of inquiry zero in on the
dynamics of Tor and the Dark Web that occur after the users have adopted the technol-
ogy. Comparatively little attention has been paid to what happens before the actual
usage, which is a fundamental question to understand how ordinary users get to know
and finally use the technology.
Contextual factors, such as the level of political repression in society, apparently show
significant correlations with the total usage of the Tor network (Jardine, 2018b). However,
even if situational features have incentivized the use of Tor, actual usage could hardly be
accomplished without sufficient awareness of the technology and enough understanding
of how to use it. Overall, the research gaps in the current literature beg the following
research question: Does a knowledge accumulation process precede the usage of the Tor
network?
The current study aims to understand the global usage of the Tor network by examin-
ing the facets of the knowledge accumulation process. We conjecture that the knowledge
accumulation process that guides users into using a media technology may consist of
three discrete steps: being motivated, seeking knowledge, and eventual use. There are
two types of knowledge, broad awareness-knowledge and specific how-to knowledge
(Rogers, 2003), in the knowledge accumulation process, which promotes and enhances
the use of the Tor network. First, broad awareness-knowledge, such as a basic under-
standing of the socio-political community dynamics on the Onion/Hidden Services, can
lead directly to the usage of Tor. Second, functional how-to knowledge, such as knowing
how to download and install the Tor browser, can lead directly to Tor use and also be a
necessary precondition of use after being aware of the technology.
Chen et al. 3

In this study, we first describe the role of knowledge in using new media technology
from the theoretical perspective of the innovation-decision process model (Rogers, 2003)
and then propose research hypotheses. Next, we operationalize the knowledge accumu-
lation process via the aggregated volume of Google searches for multiple topics. Country-
level panel data from 2013 to 2017 are collected for testing the research hypotheses with
linear mixed-effects models. Both direct and indirect relationships between the knowl-
edge accumulation process and global usage of the Tor network are revealed. Finally, we
examine how the level of freedom conditions the relationships between the knowledge
accumulation process and eventual usage under different political contexts.

Knowledge accumulation in the use of new media technology

Before the eventual use of new media technology, contextual settings are the predomi-
nant motivators before any action is taken. Prior contextual motivators usually include
multiple factors, from individual-level factors, such as personal needs, to societal-level
factors, such as the norms of the social system (Rogers, 2003: 170). From the perspective
of the uses and gratifications theory (Katz et al., 1974), social and psychological needs
drive the usage of different media technologies to obtain gratifications. For example,
perceived needs have been empirically supported as important predictors of the adoption
and use of the Internet (Zhu and He, 2002). Therefore, the use of new media technology
is inherently goal-oriented and motivated by background contextual factors.
The decision on the use of new media technology is not an instantaneous act but rather
a process with a sequence of actions, where knowledge is an indispensable part. Based
on the model of the innovation-decision process, there are five stages through which a
user moves before eventual usage—knowledge, persuasion, decision, implementation,
and confirmation (Rogers, 2003). Since the whole model is a linear and sequential pro-
cess, the first stage, knowledge, plays an extremely important role. Only after being
equipped with a considerable amount of knowledge about the technology can a potential
user possibly move through the further stages of the innovation-decision process.
Empirical evidence has confirmed that knowledge accumulation matters for the use of
new media technology, particularly in digital domains, such as protective information
technologies (Dinev and Hu, 2007), smart card technology (Taherdoost and Masrom,
2009), and mobile payment platforms (Cao and Niu, 2019). Although definitive empirical
evidence is difficult to capture on every discrete stage of the innovation-decision process,
as it is difficult to clearly distinguish the intrapersonal mental processes of individual
respondents, the earlier (knowledge) and later (decision) stages tend to be more clearly
distinguished than all the other intervening points, such as the persuasion stage (see
Generalization 5–12 in Rogers, 2003: 198). Overall, knowledge plays a fundamental role
throughout the whole innovation-decision process.
This article focuses on the knowledge accumulation process between contextual factors and
eventual usage, which includes two types of knowledge: awareness-knowledge (what the tech-
nology is) and how-to knowledge (how to use the technology). At the very beginning, people
who are motivated to adopt a new media technology must discover the existence of such tech-
nology, that is, awareness-knowledge. After initial awareness is gained, people may further
accumulate in-depth knowledge to reveal the functional utility of the technology, particularly
4 new media & society 00(0)

Movated by Awareness-Knowledge & Eventual Use of New

Contextual Factors How-to Knowledge Media Technology

Knowledge Accumulaon Process

Figure 1. Conceptual diagram.

the details about how to employ the technology correctly, that is, how-to knowledge. The con-
ceptual diagram of the knowledge accumulation process is illustrated in Figure 1.

Knowledge accumulation in the use of the Tor network. Getting from motivation to eventual
usage requires a knowledge accumulation process, particularly for a technologically
sophisticated medium like Tor, which, together with the “invisible” Dark Web, is still not
yet widely known by the public (Devine et al., 2015). Knowledge related to Tor could be
accumulated either through active knowledge seeking, such as Google searches, or passive
exposure to various online content. Popular social media platforms, such as Facebook,
Instagram, and Reddit, as well as forums, blog posts, and YouTube videos, are all important
sources where motivated users get exposed to related knowledge about the Tor network,
either intentionally or unintentionally (Bakken and Demant, 2019; Demant et al., 2019;
Hazel Kwon and Shao, 2020). However, social media posts about Tor and the Dark Web
may be largely segregated and shared only within subcommunities (Bancroft et al., 2020),
as they may include sensitive information, such as personal experiences. Therefore, extra
efforts are still needed through active seeking to accumulate the related knowledge.
In this study, we see the use of Tor as a highly motivated process and focus on the
active searching behavior on Google as a proxy for the knowledge accumulation process
toward using Tor. Google searches have been a popular way for users to accumulate and
aggregate knowledge from various online sources (Slechten et al., 2021), and they have
been widely employed as an effective predictor of important social outcomes (Jardine
and Lindner, 2020; Nuti et al., 2014; Perdue et al., 2018). Therefore, the worldwide
search queries collected unobtrusively by Google Trends work as valid observations of
people’s knowledge accumulation processes.
The Tor network works as an anonymity-granting technology for online browsing, as
well as other social interactions, by protecting a user’s online identity and personal privacy
(Bancroft and Reid, 2017; Sotirakopoulos, 2018). A worldwide survey has demonstrated
that concerns about privacy, censorship, and data breaches significantly predict attitudes
toward the Tor network (Jardine, 2018a). Subnational studies further found that people
from metropolitan areas with more liberal citizen ideology are more likely to become inter-
ested in Tor and search for related knowledge (Lindner et al., 2020). In particular, existing
evidence in the United States at a state-year level has suggested that searches for topics
related to Edward Snowden’s revelations are more potent for predicting the search popular-
ity of Tor in comparison to searches for the Dark Web (Lindner and Xiao, 2020). Therefore,
people with deep concerns about online privacy and mass surveillance are likely to search
for protective solutions, which may help accumulate knowledge of the Tor network.
Chen et al. 5

The highly anonymous feature of the Tor network has also created a perfect platform
for underground activities and malicious use (Maddox et al., 2015; Rhumorbarbe et al.,
2018). While legal content exists in high proportions on the Tor network (Faizan and
Khan, 2019; Moore and Rid, 2016), many of the Onion/Hidden services run cryptomar-
kets that provide global access to illicit substances, particularly drugs (Christin, 2013;
Martin, 2014). Furthermore, drug websites turn out to be the most frequently visited
Onion/Hidden services on the Tor network (Owen and Savage, 2016). Also, a subna-
tional study showed that search interests on the Dark Web significantly predict the con-
sumption of cannabis in the United States (Jardine and Lindner, 2020). Therefore, people
who search for topics related to online drugs dealings and the Dark Web markets could
also possibly accumulate knowledge of the Tor network.
As the Tor network is specially designed and technically sophisticated, the how-to
knowledge is important even after a potential user has already gained awareness of the
technology. A motivated user of the Tor network must still undergo a nontrivial process
of knowledge accumulation to learn about, navigate, find, download, install, and effec-
tively use the Tor browser to access the Tor network for their intended purpose.
Furthermore, some supporting techniques may help facilitate the use of the Tor network
more safely, such as VPN, Tails, and PGP (Kowalski et al., 2019). Using these technolo-
gies together would be much more secure than simply using the Tor network on its own.

Mechanisms of the knowledge accumulation process. In considering the utilities of the Tor
network (Jardine et al., 2020), we identified two clusters of topics that could capture the
awareness-knowledge accumulation process of potentially motivated Tor users: “Digital
Rights” and “Drug Markets” (see Table 1 in the “Method” section for detailed topics of each
cluster). “Digital Rights” mainly refers to existing human rights, such as the right to privacy
and freedom of expression, in the context of digital technologies, particularly the Internet.
According to the United Nations, “the same rights that people have offline must also be
protected online” (Human Rights Council UN, 2016). Digital rights allow individuals to
maintain their privacy and freedom of expression when using the Internet. In addition to
“Digital Rights,” people who are interested in the “Drug Markets” are also likely to be inter-
ested in Tor. The “Drug Markets” cluster broadly refers to topics related to the whole eco-
system for drug dealing on the Tor network, particularly the Dark Web. When considering
the two clusters’ awareness-knowledge and how-to knowledge together, three mechanisms
depict how knowledge accumulation leads to the eventual usage of the Tor network.

From awareness-knowledge to how-to knowledge. People usually need to be fully aware

of a technology before searching for it to gain how-to knowledge. The search terms
readily reveal the degree of interest that people have in potential uses of the technol-
ogy, which may incentivize further accumulation of how-to knowledge. For example,
evidence from a cross-national analysis has shown that perceived concern about pri-
vacy and exposure to certain motivated offenders could effectively lead to target-hard-
ening behaviors, that is, pursuing privacy-granting technologies, such as two-factor
authentication and antivirus software (Ireland, 2021). Therefore, these conceptual
steps leading from the search for awareness-knowledge to how-to knowledge suggest
the following hypotheses:
6 new media & society 00(0)

Table 1. Search topics in each type of knowledge accumulation.

Type of knowledge accumulation Search topics References

“Tor”-related knowledge Tor Lindner and Xiao (2020)
accumulation (M = 1.51, SD = 2.70) Tails Kowalski et al. (2019)
PGP Kowalski et al. (2019)
“Digital Rights”-related knowledge Surveillance Lindner and Xiao (2020)
accumulation (M = 3.03, SD = 4.83) Censorship Jardine (2018a)
Privacy Jardine (2018a)
Private browsing Jardine (2018a)
Data breach Jardine (2018a)
“Drug Markets”-related knowledge Dark Web Lindner and Xiao (2020)
accumulation (M = 3.30, SD = 5.60) Deep Web Lindner and Xiao (2020)
Silk Road Lindner and Xiao (2020)
Heroin Tzanetakis (2018)
THC Van Buskirk et al. (2016)
MDMA Tzanetakis (2018); Van Buskirk et al.
(2016)
Cocaine Tzanetakis (2018); Van Buskirk et al.
(2016)

H1a: Google searches for “Digital Rights”-related awareness-knowledge will be posi-

tively correlated with Google searches for “Tor”-related how-to knowledge.
H1b: Google searches for “Drug Markets”-related awareness-knowledge will be posi-
tively correlated with Google searches for “Tor”-related how-to knowledge.

From how-to knowledge to eventual use. Before using the technology, people must be well-
informed about how to use the technology, as Tor is technologically sophisticated. Learn-
ing how-to knowledge is particularly important for complex innovations in the
innovation-decision process (Rogers, 2003: 173). If an adequate level of how-to knowl-
edge cannot be acquired before the trial and adoption, rejection and discontinuance are
quite likely to happen, which in turn decreases the level of usage. Therefore, the follow-
ing research hypothesis is proposed:

H2: Google searches for “Tor”-related how-to knowledge will be positively associ-
ated with daily Tor usage level.

Since awareness-knowledge tends to correlate with how-to knowledge, and how-to

knowledge is the fundamental variable associated with the outcome of the innovation-
decision process, it is reasonable to infer indirect relationships between awareness-
knowledge and usage of the Tor network through how-to knowledge accumulation. After
being exposed to “Digital Rights” and “Drug Markets”-related information about Tor via
Google searches, users might search for “Tor”-related topics to learn how to use the
technology. In other words, the acquisition of how-to knowledge could mediate users’
Chen et al. 7

eventual use after they get to know it. Therefore, we propose the following research
hypotheses:

H2a: Google searches for “Digital Rights”-related awareness-knowledge will have a

positive indirect relationship on daily Tor usage level via Google searches for “Tor”-
related how-to knowledge.
H2b: Google searches for “Drug Markets”-related awareness-knowledge will have a
positive indirect relationship on daily Tor usage level via Google searches for “Tor”-
related how-to knowledge.

From awareness-knowledge to eventual use. To fully understand the nature of the knowledge
accumulation process and how it works to influence eventual adoption, it is essential to
hold a dynamic perspective (Rogers, 2003: 196). The distinct types of knowledge accumu-
lation could vary as a function of the dynamism of the technology. Before the first use of
the technology, how-to knowledge plays an indispensable role in the adoption process, as
it demonstrates how the technology works for the potential adoptions. However, once a
technology is already in use, the accumulation of additional functional knowledge tends to
have limited utility. Since the whole ecosystem of the Tor network is highly unpredictable,
with old sites vanishing while new sites emerge (Owenson et al., 2018), it is still necessary
to keep the awareness-knowledge refreshed and updated to sustainably use the technology,
even for experienced users. The emergence of subcommunity engagement further suggests
that people must become routinized to participate continuously (Gehl, 2016, 2018; Munks-
gaard and Demant, 2016). As the search volume for both “Digital Rights”- and “Drug
Markets”-related topics also acts as a good proxy for the refreshing process of sustaining
awareness-knowledge related to the Tor network, the following hypotheses are proposed:

H3a: Google searches for “Digital Rights”-related awareness-knowledge will be posi-

tively correlated with daily Tor usage level.
H3b: Google searches for “Drug Markets”-related awareness-knowledge will be posi-
tively correlated with daily Tor usage level.

Country differences in the knowledge accumulation process. The knowledge accumulation

process along the path from awareness-knowledge to how-to knowledge could vary
between people from different social and political contexts. People from less free regimes
may be faced with more political needs to seek online anonymity for censorship circum-
vention compared with users from free countries (Jardine, 2018b). Therefore, people
from non-free countries tend to be more motivated than those from free regimes to take
a further step to search for “Tor”-related how-to knowledge when they are already aware
of the technology. The following research hypothesis is proposed:

H4a: Relative to non-free regimes, Google searches for “Digital Rights”-related

awareness-knowledge will show a weaker correlation with how-to knowledge
searches in free regimes.
8 new media & society 00(0)

Drug-trading activities are usually clustered in specific countries, such as the

United States, United Kingdom, Australia, China, and the Netherlands (Van Buskirk
et al., 2016). Traffic analysis shows that free countries tend to have a higher per-
centage of users who access Onion/Hidden Services (i.e. the Dark Web, such as
cryptomarkets for drug trading) than non-free countries (Jardine et al., 2020).
Therefore, it is more likely for users in free regimes to search for Tor, driven by
interests in the drug markets, than those in non-free regimes. The following research
hypothesis is proposed:

H4b: Relative to non-free regimes, Google searches for “Drug Markets”-related

awareness-knowledge will show a stronger correlation with how-to knowledge
searches in free regimes.

In addition to searching for “Tor”-related how-to knowledge, country differences are

also likely to be observed in the eventual usage, as the information environment might be
highly dependent on sociopolitical systems. People from less free regimes are more
likely to be influenced by information control measures, such as information frictions,
which refer to governments or interest groups making information more disaggregated
and more inaccessible to the public (Roberts, 2014). Information frictions not only
decrease the accessibility of information but also disaggregate the story of a particular
event and make it hard to bring different pieces of information together. Consequently,
the spread of online information will be slower, and it is less unlikely for many Internet
users to become aware of a particular event in less free countries (Roberts, 2014).
The Tor network, as a highly effective tool for censorship circumvention, is particu-
larly likely to be influenced by information frictions. For example, rigorous censorship
tools have been developed to fully block the Tor network (Dunna et al., 2018). Compared
with people from less free regimes, users from more free countries will have more oppor-
tunities to gain access to related knowledge and fewer concerns about state counterac-
tions to employ privacy-enhancing technologies (Jardine, 2018b). Therefore, the
following hypothesis is proposed:

H5a: Relative to non-free regimes, Google searches for “Digital Rights”-related

awareness-knowledge will show a stronger correlation with daily Tor usage levels in
free regimes.

For “Drug Markets”-related awareness-knowledge, information frictions in less

free countries tend to have even greater deterrent effects for both new and experienced
users, as the underground activities on the Dark Web are even more likely to be illegal
and lead to arrests. Consistent with this idea, empirical evidence suggests that the
activities on the Dark Web cluster in more free regimes rather than less free regimes
(Jardine et al., 2020), which implies that the accumulation of “Drug Markets”-related
awareness-knowledge may be more likely to promote Tor usage in more free countries.
Therefore, the following research hypothesis is proposed (all the research hypotheses
are illustrated in Figure 2):
Chen et al. 9

Google Searches for

“Digital Rights”
Awareness-Knowledge

H1a

H4a H5a H3a

Google Searches for

Level of Freedom “Tor” Related Daily Tor Usage Level
How-to Knowledge H2

H4b H5b H3b

H1b

Google Searches for

“Drug Markets”
Awareness-Knowledge

Figure 2. Research hypotheses.

H5b: Relative to non-free regimes, Google searches for “Drug Markets”-related

awareness-knowledge will show a stronger correlation with daily Tor usage levels in
free regimes.

Method
Tor usage
There are two ways of accessing the Tor network: bridges or relays. Relays are the build-
ing blocks of the Tor network, and they serve as nodes in the overlay network. Bridges are
non-public addresses designed to help users from repressive regimes that have imposed
Internet restrictions on accessing the Tor network. Bridges connect the restricted area and
the regular relay system of the Tor network. The number of Tor relay users and bridge
users can be retrieved from Tor Metrics (https://metrics.torproject.org/), an official sub-
project for monitoring volumetric activity on the Tor network. The data measure distinct
client sessions, not unique individual users. In our analysis, we aggregate both bridge and
relay users to measure the total volume of Tor clients per country year. Tor provides sta-
tistics from 6 March 2011, to the present. These data have been widely used in published
research to characterize the geographic distribution of Tor users around the world on the
Tor network, for example, Jardine (2018b) and “The Anonymous Internet” project by the
Oxford Internet Institute. We further take the daily averaged usage (divided by 365,
M = 8775.91, SD = 33,292.16) as the eventual outcome variable in the regression models.

Google searches
Online search data were retrieved from Google Trends. We employed the volume of
“search topics” in Google Trends, defined as a bundle of correlated terms that includes
10 new media & society 00(0)

all potentially related terms across different languages. For example, the search terms
under the search topic “Censorship” include censorship (English), censura (Spanish),
цензуры (Russian), ‫( سانسور‬Persian), 審査 (Japanese), and so on. Although the exact
search terms are not disclosed, search topics in Google Trends are generally considered
to provide more comprehensive information compared with single search terms (Lindner
et al., 2020). To systematically examine the potential related topics to Tor, we compiled
terms from various previous studies for three types of knowledge accumulation (see
Table 1), that is, the accumulation of “Tor”-related knowledge, “Digital Rights”-related
knowledge, and “Drug Markets”-related knowledge, respectively.
“Tor”-related knowledge accumulation involved search topics on Tor, as well as on
other highly related technologies, such as Tails and PGP, which provides higher-level
security when using Tor (Kowalski et al., 2019). “Digital Rights”-related knowledge
accumulation included search topics on “Surveillance” (Lindner et al., 2020; Lindner
and Xiao, 2020) and “Censorship” to refer to the user interests of mass surveillance by
governments and institutions, which are important factors that determine attitudes toward
Tor (Jardine, 2018a). This cluster also included “Privacy,” “Private browsing,” and “Data
breach” to cover users’ concerns about personal data leakage in the online environment,
which are also significant drivers of using Tor (Jardine, 2018a). Finally, the “Drug
Markets” cluster included search topics that covered the whole drug-trading ecosystem
on the Internet, from specific drug names that prevail on the Dark Web marketplaces,
such as heroin, THC, MDMA, cocaine (Tzanetakis, 2018; Van Buskirk et al., 2016), and
the globally well-known “Silk Road” market to the general terms “Dark Web” and “Deep
Web” (Lindner and Xiao, 2020), which are sheltering worldwide underground drug deal-
ings. Under each type of knowledge accumulation, the search volumes were averaged
and merged into a single index. The indices were further normalized across country and
year (see Supplementary Materials for technical details). The mean and standard devia-
tions of the eventual search indices of the three topic clusters (normalized across country
and year) are shown in Table 1.

Analytical approach
As time is an indispensable element in the diffusion process of innovative technology
(Rogers, 2003), the data were compiled to an unbalanced panel for each country (N = 247),
with repeated measures (from 2013 to 2017) for longitudinal analysis.1 We took the year
as time units to focus on the macrolevel associations between search volume and average
daily usage of Tor across countries. Monthly observations could introduce more informa-
tion but may also be sensitive to noises. For example, censorship events could lead to
drastic short-term changes to Tor usage (see https://gitlab.torproject.org/tpo/network-
health/metrics/timeline for the event list). Given the panel structure of the data, we
employed linear mixed-effect models. The models included both random effects for each
country and the fixed effects of country-level control variables. Four country-level vari-
ables were included to control for country-specific variations in social and political con-
texts (see Table 2). In the regression models, the outcome variables were logged due to
the high skewness. The Q–Q plots of residuals in the regression models are shown in the
Supplementary Materials.
Chen et al. 11

Table 2. Country-level variables.

Variable name Description Source

Level of freedom Overall freedom status: non-free Freedom House
(NF), partially free (PF), and free (F)
Internet penetration Individuals using the Internet (% of World Development Indicators
(M = 48.72, SD = 28.91) population)
Population Number of residents (logged due to World Development Indicators
(M = 15.24, SD = 2.41) high skewness)
Economic openness The ratio of international trade World Development Indicators
(M = 94.51, SD = 58.70) volume in GDP
GDP per capita Dropped in the statistical analyses World Development Indicators
due to high correlation with Internet
penetration rate (r > .8)

Figure 3. Daily Tor usage over the world (averaged from 2013 to 2017).

Results
Descriptive analysis
To illustrate the geographical distribution of Tor usage worldwide, a world map of daily
Tor usage by country was created using the averaged number collected from 2013 to
2017 (see Figure 3). The map demonstrates that the global usage of the Tor network is
unequally distributed. Usage from free countries in Europe, North America, and Australia,
as well as less free countries such as Russia, Iran, and Saudi Arabia, are among the high-
est in the world. The general pattern revealed from the world map created by our col-
lected data is consistent with the cartogram made by “The Anonymous Internet” project
(http://geography.oii.ox.ac.uk/the-anonymous-internet/) in 2014.
Figure 4 shows the Tor usage and search topics in Google Trends over time, where
“NF” stands for non-free countries, “PF” stands for partially free countries, and “F”
stands for free countries. The y-axis denotes the value of the two outcome variables and
 12

Figure 4. Longitudinal trends of the key variables from 2013 to 2017.

new media & society 00(0)
Chen et al. 13

the two explanatory variables. In the two left subplots, the y-axes are the logged number
of daily Tor usage and the logged value of the Tor-related search index (normalized by
country and year), respectively. In the two right subplots, the y-axes are the indices of
Google searches for “Digital Rights” and “Drug Markets” (normalized by country and
year). All the yearly trends are more popular in more free countries than in less free
countries. Since 2014, there has been an evident decrease in Tor usage after an increase
in 2013. This is probably related to a steep increase in users due to Snowden’s 2013 dis-
closure, which had a “chilling effect” (Rosso et al., 2020), and then the number of users
declined to a relatively stable level (Hampson and Jardine, 2017). The trend of Google
searches for “Tor”-related knowledge is consistent with the trends of Tor usage.

Hypotheses testing
With all control variables included (see Model 2 in Table 3), Google searches for “Digital
Rights”-related knowledge were positively associated with Google searches for “Tor”-
related knowledge (B = 0.024, 95% CI = [0.012, 0.037], p < .001). Google searches for
“Drug Markets”-related knowledge were positively associated with Google searches for
“Tor”-related knowledge (B = 0.031, 95% CI = [0.018, 0.043], p < .001). The results were
the same with no control variables (see Model 1 in Table 3). Overall, both H1a and H1b
are supported.
With all predictors and control variables included (see Model 7 in Table 4), Google
searches for “Tor”-related knowledge were positively correlated with daily Tor usage
level (B = 0.034, 95% CI = [0.015, 0.053], p < .001). Therefore, H2 is supported.
Statistical tests were conducted on the indirect effects using the mediation package in
R. For Google searches for “Digital Rights”-related knowledge on daily Tor usage level
via Google searches for “Tor”-related knowledge, the average causal mediation effect
(ACME) was 0.004 and significant (95% CI = [0.000, 0.00], p < .001), the average direct
effect (ADE) was 0.048 and significant (95% CI = [0.029, 0.07], p < .001), and the total
effect was 0.049 and significant (95% CI = [0.029, 0.07], p < .001). The averaged indi-
rect effect was significantly positive yet relatively small. Therefore, H2a is supported.
For Google searches for “Drug Markets”-related knowledge on daily Tor usage level via
Google searches for “Tor”-related knowledge, the ACME was 0.001 and significant
(95% CI = [0.000, 0.00], p < .001), the ADE was 0.040 and significant (95% CI = [0.020,
0.06], p < .001), and the total effect was 0.041 and significant (95% CI = [0.020, 0.06],
p < .001). The averaged indirect effect was significantly positive yet relatively small.
Therefore, H2b is supported.
With all predictors and control variables included (see Model 7 in Table 4), Google
searches for “Digital Rights”-related knowledge were positively associated with daily
Tor usage level (B = 0.048, 95% CI = [0.029, 0.067], p < .001). Google searches for
“Drug Markets”-related knowledge were positively associated with daily Tor usage level
(B = 0.040, 95% CI = [0.020, 0.060], p < .001). The results were the same when there
were no control variables and without Google searches for “Tor”-related knowledge (see
Models 5 and 6 in Table 4). Therefore, both H3a and H3b are supported.
Furthermore, the correlation between Google searches for “Digital Rights”-related
knowledge and Google searches for “Tor”-related knowledge was conditioned by the
 14

Table 3. Linear mixed-effects models predicting Google searches for Tor (logged).
Predictors Model 1 Model 2 Model 3 Model 4

Estimates Estimates Estimates Estimates

(Intercept) .209*** (0.143–0.276) −2.432*** (−3.151 to −1.713) −2.456*** (−3.167 to −1.745) −2.450*** (−3.167 to −1.732)
Google searches for Digital Rights .046*** (0.036–0.057) .024*** (0.012–0.037) .050*** (0.027–0.073) .024*** (0.011–0.036)
Google searches for Drug Markets .058*** (0.048–0.068) .031*** (0.018–0.043) .033*** (0.021–0.046) .045 (−0.008 to 0.097)
Level of freedom factor (F) .173* (0.001–0.344) .239* (0.052–0.425) .204* (0.019–0.390)
Level of freedom factor (PF) .020 (−0.116 to 0.156) .095 (−0.054 to 0.243) .021 (−0.124 to 0.167)
Internet penetration .009*** (0.006–0.011) .008*** (0.006–0.011) .009*** (0.006–0.011)
Population (logged) .159*** (0.117–0.200) .157*** (0.116–0.199) .159*** (0.118 0.201)
Economic openness .001 (−0.001 to 0.002) .001 (−0.001 to 0.002) .001 (−0.001 to 0.002)
Year −.046*** (−0.066 to −0.026) −.048*** (−0.068 to −0.028) −.046*** (−0.066 to −0.026)
Google searches for Digital −.031* (−0.059 to −0.004)
Rights × level of freedom factor (F)
Google Searches for Digital −.034* (−0.061 to −0.007)
Rights × level of freedom factor (PF)
Google searches for Drug −.016 (−0.068 to 0.036)
Markets × level of freedom factor (F)
Google Searches for Drug −.009 (−0.063 to 0.046)
Markets × level of freedom factor (PF)
Random effects
σ2 0.11 0.13 0.13 0.13
τ00 0.18Country 0.16Country 0.15Country 0.15Country
ICC 0.62 0.54 0.53 0.54
N 236Country 177Country 177Country 177Country
Observations 1166 870 870 870
Marginal R2/conditional R2 .468/.799 .584/.810 .591/.810 .586/.808

*p < .05, **p < .01, ***p < .001.

new media & society 00(0)
Table 4. Linear mixed-effects models predicting daily tor usage (logged).
Chen et al.

Predictors Model 5 Model 6 Model 7 Model 8 Model 9

Estimates Estimates Estimates Estimates Estimates

(Intercept) 5.905*** (5.591–6.218) −6.536*** (−7.950 to −5.121) −6.395*** (−7.803 to −4.986) −6.340*** (−7.772 to −4.908) −6.321*** (−7.768 to −4.873)
Google searches for Digital Rights .054*** (0.037–0.072) .051*** (0.032–0.070) .048*** (0.029–0.067) .024 (−0.014 to 0.062) .049*** (0.030–0.068)
Google searches for Drug Markets .051*** (0.030–0.072) .044*** (0.024–0.064) .040*** (0.020–0.060) .039*** (0.019–0.059) −.045 (−0.124 to 0.035)
Level of freedom factor (F) .594*** (0.289–0.900) .589*** (0.286–0.893) .530** (0.199–0.861) .476** (0.148–0.803)
Level of freedom factor (PF) .278* (0.056–0.500) .277* (0.057–0.498) .216 (−0.022 to 0.454) .220 (−0.011 to 0.451)
Internet penetration .028*** (0.024–0.033) .027*** (0.023–0.032) .027*** (0.022–0.032) .027*** (0.022–0.032)
Population (logged) .728*** (0.646–0.811) .717*** (0.635–0.800) .718*** (0.634–0.801) .719*** (0.634–0.803)
Economic openness .003** (0.001–0.006) .004** (0.001–0.006) .004** (0.001–0.006) .004** (0.001–0.006)
Year −.114*** (−0.142 to −0.085) −.105*** (−0.134 to −0.076) −.101*** (−0.130 to −0.072) −.102*** (−0.131 to −0.074)
Google searches for Tor .034*** (0.015–0.053) .035*** (0.016–0.054) .032*** (0.013–0.051)
Google searches for Digital .028(−0.017–0.072)
Rights × level of freedom factor (F)
Google searches for Digital .030 (−0.012 to 0.073)
Rights × level of freedom factor (PF)
Google searches for Drug .091 *(0.010–0.172)
Markets × Level of Freedom Factor (F)
Google searches for Drug .081 (−0.002–0.165)
Markets × level of freedom factor (PF)
Random effects
σ2 .19 .23 .22 .22 .22
τ00 5.61Country 0.80Country 0.79Country 0.82Country 0.85Country
ICC 0.97 0.78 0.78 0.79 0.79
N 236Country 177Country 177Country 177Country 177Country
Observations 1166 870 870 870 870
Marginal R2/conditional R2 .042/.968 .789/.953 .791/.954 .784/.954 .780/.955

*p < .05, **p < .01, ***p < .001.

15
16 new media & society 00(0)

Google Searches for Tor (Logged)

2.0

Level of Freedom
1.5
F

PF
1.0
NF

0.5

0 5 10 15 20 25
Google Searches for Digital Rights

Figure 5. Interaction plots.

level of freedom in the country (see Model 3 in Table 3). In free and partially free coun-
tries, the associations were both significantly weaker than those in non-free countries
(see Figure 5). In contrast, the interaction term between Google searches for “Drug
Markets”-related knowledge and level of freedom was not significant (see Model 4 in
Table 3). Therefore, H4a is supported while H4b is not supported.
The correlation between Google searches for “Drug Markets”-related knowledge and
daily Tor usage level was also conditioned by the level of freedom (see Model 8 in Table
4). The correlation in non-free countries was significantly weaker (even negative) than
in free countries (see Figure 6). In other words, when there are more Google searches for
“Drug Markets”-related knowledge, there will be a lower level of daily usage of the Tor
network in non-free countries while a higher level of daily usage of the Tor network in
free countries. However, the interaction term between Google search for “Digital
Rights”-related knowledge and level of freedom was not significant (see Model 9 in
Table 4). Therefore, H5a is not supported, while H5b is supported.
In sum, both H1a and H1b are supported. H2, H2a, and H2b are all supported. H3a and
H3b are both supported. H4a is supported, while H4b is not supported. H5a is not sup-
ported, while H5b is supported. All the hypotheses testing results are presented in Figure 7.

Discussion
First, the analytical results show that Google searches for “Digital Rights” and “Drug
Markets”-related knowledge are both positively associated with daily Tor usage, while
the correlation is stronger for “Digital Rights”-related than “Drug Markets”-related
searches. The results echo a previous study which found that more than 90% of Tor users
access the Clear Web through the Tor network for web browsing to protect their privacy
from state surveillance (Jardine et al., 2020). Consistent with the highly unequal propor-
tions of the two types of activities, the findings clearly show that concerns about
Chen et al. 17

Daily Tor Usage (Logged)

8

Level of Freedom

6 F

PF

NF
4

0 10 20 30
Google Searches for Drug Markets

Figure 6. Interaction plots.

Google Searches for

“Digital Rights”
Awareness-Knowledge
0.024***
0.048***
(-) (n.s.)
(origin: 0.049***)

Google Searches for

Level of Freedom “Tor” Related Daily Tor Usage Level
How-to Knowledge 0.034***

(n.s.) (+) 0.040***

(origin: 0.041***)
0.031***

Google Searches for

“Drug Markets”
Awareness-Knowledge

Figure 7. Path diagram.

surveillance and censorship are much more important than interests in illicit drugs and
the Dark Web in motivating the use of the Tor network.
Second, Google searches for both “Digital Rights” and “Drug Markets”-related knowl-
edge are also positively associated with “Tor”-related searches, and those “Tor”-related
searches are further positively correlated to daily Tor usage. The significant yet weak
indirect associations between awareness-knowledge and daily Tor usage level via how-to
knowledge suggest that it is likely to see a two-step knowledge accumulation process for
potential new Tor users before their eventual usage of the technology. Nevertheless, the
effect size is small, suggesting that this mechanism may be relatively weak.
18 new media & society 00(0)

Furthermore, the correlation between Google searches for “Digital Rights”-related

knowledge and “Tor”-related knowledge is weaker both in free and partially free coun-
tries than in non-free countries. These results suggest that motivated users in non-free
countries are more likely than in both free and partially free countries to search for “Tor”-
related knowledge when they search for “Digital Rights”-related knowledge. Meanwhile,
the correlation between Google searches for “Drug Markets”-related knowledge and
daily Tor usage level is significantly stronger in free countries compared with non-free
countries. The results suggest that Tor users in free countries are more likely to be moti-
vated by interests in Dark Web activities in free regimes where information frictions are
weaker, while users in non-free countries may be deterred by various political restric-
tions and therefore give up using the technology due to information frictions.
This study provides a thorough understanding of the knowledge accumulation process
leading to the global usage of a sophisticated dual-use technology—the Tor network.
Employing linear mixed-effects models with country-year panel data, our analyses dem-
onstrate that the knowledge accumulation process plays an essential role in guiding moti-
vated users into eventual use. There are two major mechanisms in the knowledge
accumulation process. First, we observe a direct relationship between the active search
for awareness-knowledge, motivated by the suppression of digital rights or the need for
illicit commodities, and eventual use. We conjecture that this mechanism could depict
the usage pattern of experienced Tor users. Second, the active seeking for how to use the
technology could bridge the awareness and eventual use of the technology. We conjec-
ture that this mechanism could describe the usage pattern of new Tor users.

Contributions
The major contributions of this study are twofold. First, the study fills in the existing research
gap between contextual motivators and eventual use of new media technology by examining
the predictors of Tor usage levels from the perspective of the knowledge accumulation pro-
cess. Furthermore, existing studies have only examined searching behavior as a predictor of
the search for “Tor”-related topics or offline activities within the United States (Jardine and
Lindner, 2020; Lindner et al., 2020; Lindner and Xiao, 2020). This study extends the out-
come variable from the search for “Tor” to real usage in a cross-national setting. Second,
country differences are found in the associations between knowledge accumulation and
eventual usage of Tor by the level of freedom. Motivated users in non-free regimes tend to
be more likely to seek out Tor to overcome censorship, while the real usage is still clustered
in free countries, largely driven by interests in Dark Web activities.
For policymakers, this study helps more clearly identify the potential motivations for
using the Tor network through a worldwide investigation of collective attention. Although
the Dark Web has become popularly known for its facilitation of illegal trading activities,
which is one of the widely known purposes of using the Tor network, “many users of the
Darknet might not perceive it as intrinsically criminogenic” (Mirea et al., 2019). The Tor
network can be utilized for good and “legitimate” purposes (Gehl, 2018), such as protect-
ing political dissidents and freedom of speech. Therefore, this study has provided addi-
tional empirical evidence and insights into the “Dark Web Dilemma” (Jardine, 2015) in
that the Tor network is a “dual-use technology” that can be used for awful purposes as
Chen et al. 19

well as for good, where the boundaries are difficult for the law to clearly define (Gehl,
2018).

Limitations and future work

It is necessary to acknowledge the limitations of the current study, which largely
come from the unit of analysis. The research question we tried to answer in this arti-
cle concerned how users become aware of and learn to use the Tor network, which is
an individual-level process. However, due to the restrictions and limitations of data
collection, the data analysis is primarily conducted at the aggregate level. We are
fully aware of the problems related to making inferences from aggregate data (e.g.
ecological fallacy). Therefore, we suggest that the findings in this study should be
taken cautiously, particularly when trying to infer individual-level behavior.
Furthermore, due to the restrictions of Google Trends, we are not aware of what
Google Trends considers part of a topic. Therefore, it is possible to see overlaps in
search results between different search topics. For example, the Dark Web topic may
include Tor as a search term, and the Tor topic could include Dark Web as a search
query. It is also necessary to point out that the limitation is largely determined by the
nature of the Tor network. Since the technology is designed to disassociate individ-
ual identities from their online activities, there are inherent obstacles to studying Tor
user behavior in online settings.
We suggest three promising directions that future work may attempt to pursue. First,
individual-level data may be collected to further confirm and extend the findings in the
current work. For example, individuals’ perceptions and knowledge about the Tor net-
work and their actual behavior on the Tor network could be collected through surveys.
However, representative sampling could be another potential problem, as the population
of all Tor network users may be difficult to know. Second, since this study is based on a
country-year panel design, future work with individual-level data could employ a more
fine-grained design on time granularity, such as monthly or weekly. More analytical
approaches, such as an interrupted time-series analysis, could be used to reveal the influ-
ence of big events related to Tor usage across the world. The third promising future direc-
tion is to examine some other potential approaches to knowledge accumulation other than
Google searches. Many users of the Tor network will also participate in online discussions
on certain platforms on the Surface Web besides the Dark Web, such as Reddit (Hazel
Kwon and Shao, 2020) and Stack Exchange (https://tor.stackexchange.com/). In this
study, the examinations of Google searches have only captured some of the steps for the
knowledge accumulation process through search behavior. The study may provide more
insights by further examining how users discuss the technology, that is, the user-generated
content on the Internet, as well as the passive exposure to such ideas via social media.

Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/
or publication of this article: This work is supported by CityU Research Project No. 9618016 (ZC
and XFL) and SRG7005702 (JJHZ) and the Major Project of the National Social Science Fund of
China No. 19ZDA324 (XFL and JJHZ).
20 new media & society 00(0)

ORCID iDs
Zhicong Chen https://orcid.org/0000-0003-0334-9454
Eric Jardine https://orcid.org/0000-0002-2041-314X
Xiao Fan Liu https://orcid.org/0000-0001-8342-4623
Jonathan J. H. Zhu https://orcid.org/0000-0001-6173-6941

Supplemental material
Supplemental material for this article is available online.

Note
1. The observation stops at 2017 because there are too many missing values of Internet penetra-
tion in 2018 and 2019, according to the World Development Indicators provided by the World
Bank.

References
Bakken SA and Demant JJ (2019) Sellers’ risk perceptions in public and private social media drug
markets. International Journal of Drug Policy 73: 255–262.
Bancroft A and Reid PS (2017) Challenging the techno-politics of anonymity: the case of crypto-
market users. Information Communication & Society 20(4): 497–512.
Bancroft A, Squirrell T, Zaunseder A, et al. (2020) Producing trust among illicit actors: a
techno-social approach to an online illicit market. Sociological Research Online 25(3):
456–472.
Benjamin V, Valacich JS and Chen H (2019) Dice-E: a framework for conducting darknet identi-
fication, collection, evaluation with ethics. MIS Quarterly 43(1): 1–22.
Cao Q and Niu XF (2019) Integrating context-awareness and UTAUT to explain Alipay user adop-
tion. International Journal of Industrial Ergonomics 69: 9–13.
Christin N (2013) Traveling the Silk Road: a measurement analysis of a large anonymous online
marketplace. In: Proceedings of the 22nd international conference on world wide web, Rio
de Janeiro, Brazil, 13 May 2013, pp. 213–224. New York: ACM.
Decary-Hetu D and Giommoni L (2017) Do police crackdowns disrupt drug cryptomarkets? A
longitudinal analysis of the effects of Operation Onymous. Crime Law and Social Change
67(1): 55–75.
Demant J, Bakken SA, Oksanen A, et al. (2019) Drug dealing on Facebook, Snapchat and
Instagram: a qualitative analysis of novel drug markets in the Nordic countries. Drug and
Alcohol Review 38(4): 377–385.
Devine J, Egger-Sider F and Rojas A (2015) The evolving impact of the invisible web: exploring
economic and political ramifications. Journal of Web Librarianship 9(4): 145–161.
Dinev T and Hu Q (2007) The centrality of awareness in the formation of user behavioral inten-
tion toward protective information technologies. Journal of the Association for Information
Systems 8(7): 386–408.
Dingledine R, Mathewson N and Syverson P (2004) Tor: the second-generation onion router. In:
Proceedings of the 13th conference on USENIX security symposium, vol. 13, San Diego, CA,
p. 21. Berkeley, CA: USENIX Association.
Dunna A, O’Brien C and Gill P (2018) Analyzing China’s blocking of unpublished tor bridges. In: 8th
{USENIX} workshop on free and open communications on the internet {FOCI} 18), Baltimore,
MD, 14 August 2018. Berkeley, CA: USENIX Association.
Chen et al. 21

Faizan M and Khan RA (2019) Exploring and analyzing the dark Web: a new alchemy. First
Monday 24(5). https://firstmonday.org/ojs/index.php/fm/article/view/9473
Gehl RW (2016) Power/freedom on the dark web: a digital ethnography of the Dark Web Social
Network. New Media & Society 18(7): 1219–1235.
Gehl RW (2018) Weaving the Dark Web Legitimacy on Freenet, Tor, and I2P. Cambridge, MA:
The MIT Press.
Hampson FO and Jardine E (2017) Look Who’s Watching: Surveillance, Treachery and Trust
Online. Montreal, QC: McGill-Queen’s Press-MQUP.
Hazel Kwon K and Shao C (2020) Communicative constitution of illicit online trade collectives: an
exploration of Darkweb market Subreddits. In: International conference on social media and
society, Toronto, ON, Canada, pp. 65–72. New York: ACM.
Human Rights Council UN (2016) The Promotion, Protection and Enjoyment of Human Rights on
the Internet. Geneva: Human Rights Council UN.
Ireland L (2021) Predicting online target hardening behaviors: an extension of routine activity the-
ory for privacy-enhancing technologies and techniques. Deviant Behavior 42(12): 1532–1548.
Jardine E (2015) The Dark Web Dilemma: Tor, anonymity and online policing. Global Commission
on Internet Governance Paper Series. Waterloo, ON: Centre for International Governance
Innovation.
Jardine E (2018a) Privacy, censorship, data breaches and Internet freedom: the drivers of support
and opposition to Dark Web technologies. New Media & Society 20(8): 2824–2843.
Jardine E (2018b) Tor, what is it good for? Political repression and the use of online anonymity-
granting technologies. New Media & Society 20(2): 435–452.
Jardine E and Lindner AM (2020) The Dark Web and cannabis use in the United States: evidence
from a big data research design. International Journal of Drug Policy 76: 102627.
Jardine E, Lindner AM and Owenson G (2020) The potential harms of the Tor anonymity network
cluster disproportionately in free countries. Proceedings of the National Academy of Sciences
of the United States of America 117(50): 31716–31721.
Katz E, Blumler JG and Gurevitch M (1974) Uses and gratifications research. The Public Opinion
Quarterly 37(4): 509–523.
Kowalski M, Hooker C and Barratt MJ (2019) Should we smoke it for you as well? An ethno-
graphic analysis of a drug cryptomarket environment. International Journal of Drug Policy
73: 245–254.
Lindner AM, Pryciak G and Elsner J (2020) Tor and the city: MSA-level correlates of interest in
anonymous web browsing. Surveillance & Society 18(4): 507–521.
Lindner AM and Xiao TT (2020) Subverting surveillance or accessing the dark web? Interest in the
Tor anonymity network in U.S. States, 2006–2015. Social Currents 7(4): 352–370.
Maddox A (2020) Disrupting the ethnographic imaginarium: challenges of immersion in the Silk
Road cryptomarket community. Journal of Digital Social Research 2(1): 20–38.
Maddox A, Barratt MJ, Allen M, et al. (2015) Constructive activism in the dark web: cryptomar-
kets and illicit drugs in the digital “demimonde.” Information, Communication & Society
19(1): 111–126.
Martin J (2014) Drugs on the Dark Net: How Cryptomarkets Are Transforming the Global Trade
in Illicit Drugs. London: Palgrave Pivot.
Mirea M, Wang V and Jung J (2019) The not so dark side of the darknet: a qualitative study.
Security Journal 32(2): 102–118.
Moore D and Rid T (2016) Cryptopolitik and the Darknet. Survival 58(1): 7–38.
Munksgaard R and Demant J (2016) Mixing politics and crime—the prevalence and decline of
political discourse on the cryptomarket. International Journal of Drug Policy 35: 77–83.
22 new media & society 00(0)

Nuti SV, Wayda B, Ranasinghe I, et al. (2014) The use of Google Trends in health care research:
a systematic review. PLoS One 9(10): e109583.
Owen G and Savage N (2016) Empirical analysis of Tor hidden services. IET Information Security
10(3): 113–118.
Owenson G, Cortes S and Lewman A (2018) The darknet’s smaller than we thought: the life cycle
of Tor hidden services. Digital Investigation 27: 17–22.
Perdue RT, Hawdon J and Thames KM (2018) Can big data predict the rise of novel drug abuse?
Journal of Drug Issues 48(4): 508–518.
Rhumorbarbe D, Werner D, Gillieron Q, et al. (2018) Characterising the online weapons traffick-
ing on cryptomarkets. Forensic science international 283: 16–20.
Roberts ME (2014) Fear, Friction, and Flooding: Methods of Online Information Control PhD
Thesis. Harvard University, Ann Arbor.
Rogers EM (2003) Diffusion of Innovations. New York: Free Press.
Rosso M, Nasir ABM and Farhadloo M (2020) Chilling effects and the stock market response to
the Snowden revelations. New Media & Society 22(11): 1976–1995.
Slechten L, Courtois C, Coenen L, et al. (2021) Adapting the selective exposure perspective to
algorithmically governed platforms: the case of Google search. Communication Research.
Epub ahead of print 30 April 2021. DOI: 10.1177/00936502211012154.
Sotirakopoulos N (2018) Cryptomarkets as a libertarian counter-conduct of resistance. European
Journal of Social Theory 21(2): 189–206.
Taherdoost H and Masrom M (2009) An examination of smart card technology acceptance using
adoption model. In: Proceedings of the ITI 2009 31st international conference on informa-
tion technology interfaces, Cavtat, Croatia, 22–25 June 2009, pp. 329–334. New York: IEEE.
Tzanetakis M (2018) Comparing cryptomarkets for drugs. A characterisation of sellers and buyers
over time. International Journal of Drug Policy 56: 176–186.
Van Buskirk J, Bruno R, Dobbins T, et al. (2017) The recovery of online drug markets following
law enforcement and other disruptions. Drug Alcohol Dependence 173: 159–162.
Van Buskirk J, Naicker S, Roxburgh A, et al. (2016) Who sells what? Country specific differences
in substance availability on the Agora cryptomarket. International Journal of Drug Policy
35: 16–23.
Zhu JJH and He Z (2002) Perceived characteristics, perceived needs, and perceived popularity—
adoption and use of the Internet in China. Communication Research 29(4): 466–495.

Author biographies
Zhicong Chen is a Ph.D. student in the Web Mining Laboratory, Department of Media and
Communication, City University of Hong Kong. His research focuses on computational social sci-
ence and new media technology.
Eric Jardine is an Assistant Professor of political science at Virginia Tech, a Senior Fellow at the
Centre for International Governance Innovation (CIGI), and an affiliate of Concinnity Risks. His
research focuses primarily on the Dark Web and cybersecurity risk, among other topics.
Xiao Fan Liu is an Assistant Professor in the Web Mining Laboratory, Department of Media and
Communication, City University of Hong Kong. His current research focuses on the application of
computational methods in social science research, primarily in cryptocurrency and the Dark Web.
Jonathan J. H. Zhu is a Chair Professor of Computational Social Science at City University of
Hong Kong. His current research focuses on the use and impact of social and mobile media and the
application of computational methods in social science research.

View publication stats