A Digital Ethnography: View of Online content moderation and the Dark Web

This essay is an early ethnographic exploration of the Dark Web Social Network (DWSN), a social networking site only accessible to Web browsers equipped with The Onion Router. The central claim of this essay is that the DWSN is an experiment in power/freedom, an attempt to simultaneously trace, deploy, and overcome the historical conditions in which it finds itself: the generic constraints and affordances of social networking as they have been developed over the past decade by Facebook and Twitter, and the ideological constraints and affordances of public perceptions of the dark web, which hold that the dark web is useful for both taboo activities and freedom from state oppression. I trace the DWSN’s experiment with power/freedom through three practices: anonymous/social networking, the banning of child pornography, and the productive aspects of techno-elitism. I then use these practices to specify particular forms of power/freedom on the DWSN. 

Recently, published a blog post on THE DARK WEB LIST  says that anonymity is produced by disassociating a user from the content they are trying to view. The entry node might record Alice’s IP address, but it does not know the final destination of the query. The exit node might know the content that is being requested (or, minimally, the address of the site hosting the content), but not who initiated the query. While attacks on the Tor network can sometimes de-anonymize blocks of traffic (A. Johnson, et al., 2013), the system is generally robust anonymity and censorship circumvention tool with wide appeal, especially for those engaging in illegal conduct, the privacy-conscious and individuals in repressive political regimes. 

ark web ethnographic method

Because this is an ethnography of a dark web SNS, there were several methodological challenges. The dark web is different from the “clear web” in important ways. The dark web is part of the Internet that cannot be accessed by mainstream software.1 It includes hidden sites that end in “.onion” or “.i2p” or other Top-Level Domain names only available through modified browsers or special software. Accessing I2P sites requires a special routing program. Accessing non-mainstream Top-Level Domains through OpenNIC requires the user to change the DNS server addresses on his or her router. Accessing .onion sites require Tor (for a tutorial on Tor and .onions, see Hoffman (2012)).

Moreover, those who run dark websites that end in .onion are able to hide their identities and locations from most, if not all, Internet users (Dingledine et al., 2004). In most cases, a visitor to a .onion site will not know the identity of the host, nor will the host know the identity of the visitor. This is very different from the mainstream Internet, where sites are often associated with a company or location (e.g. google.com is associated with the company headquartered in Mountain View, CA), and visitors are often identified and monitored through sundry tracking technologies such as cookies, account registrations, Flash cookies, IP addresses, and geolocation.

Although these technical conditions are challenging for ethnography, they are not unique in the ethnographic literature. In doing a participant observation of the DWSN, I followed the example of anthropologist Tom Boellstorff’s (2008) ethnography of Second Life. Boellstorff takes Second Life on its own terms; he avoids linking Second Life avatars to their “real-world” counterparts in order to focus on day-to-day life in that virtual world. He treats Second Life as its own space, with its own rules and culture, rather than as articulated with the “real world” outside of the virtual. This is a methodological choice of Boellstorff’s (Winnick, 2008), as opposed to the work of, for example, Danah boyd (in press), who studies social media users both online and offline.

For me, however, I have no choice but to study the DWSN on its own terms. I cannot link DWSN avatars to their flesh-and-blood counterparts even if I wanted to. This is for the technical reasons I mention above, but also due to the culture of the DWSN. According to the DWSN privacy policy, “In order to protect everyone’s privacy, you have to protect yourself. You can do this by not giving out any personal information. No personal emails. No real names. No specific location information.” Due to this, I did not seek to learn any personal information from DWSN members (including name, age, gender, location). Moreover, due to an agreement I made with the members of the DWSN that I spoke to, all user pseudonyms have been made into new pseudonyms (i.e. “an admin” and “a member”). I did receive an exemption for this study from my university’s Institutional Review Board (IRB). Most of this material is taken from public parts of the DWSN (“public” insofar as they are accessible with Tor and a DWSN account), with some material drawn from one-on-one interviews conducted through private messaging. These interviews were preceded with informed consent notices. In all cases, in keeping with my university’s IRB standards, I offered to reveal my identity to those I interviewed, but (in keeping with the strictly anonymous character of the DWSN) my identity was refused. However, I also provided a draft of this essay to the DWSN administrators I spoke to and have revised according to their feedback (call it another round of blind peer review).

Thus, just as Boellstorff did with Second Life, I focused on how DWSN members interact within the site and how the site is structured. In this sense, I paid attention to the governance and interaction dynamics of the site in much the same manner as Geert Lovink’s (2003) analyses of the Amsterdam Digital City and nettime (but without any offline contact with DWSN members). That is, I engaged in participant observation, focusing on the intersection between site architecture (Davis, 2010; Star, 1999) and member actions in the vein of digital ethnography as explicated by Gabriella Coleman (2010), paying attention to “various frames of analysis, … history, and the local contexts and lived experiences of digital media” (p. 488).

Power/freedom: lenses to see in the dark (web)

Although the strict anonymity of the DWSN narrowed the scope of my work (in that I could not talk to DWSN members offline), an ethnography of any SNS must also be focused. Despite being less accessible than “clear web” SNSs such as Facebook or Twitter, the DWSN has shown evidence of growth since its founding in 2013. During my observation of the site over a period of 10 months, the DWSN’s number of accounts grew from 3000 to over 24,000, with over 170 groups, hundreds of blog posts, and tens of thousands of micro-blog posts. Although these are not numbers on the scale of Facebook or Twitter, they are impressive, given that finding the DWSN is not a simple matter of Googling for it. More to the point, it would be impossible to observe or speak with thousands of people.

To focus my study, I sought lenses (or “categories,” as Koopman and Matza (2013) would call them) with which to illuminate activities on the DWSN. The two I use are power and freedom. These lenses, I should stress, are not free-floating concepts that I will simply grab from a theoretician and then “apply” to the DWSN. These lenses come from the context in which the DWSN emerges: the “media ideology” (Gershon, 2010) of the dark web, specifically as this ideology appears in journalistic coverage. This coverage forms part of the historical context the DWSN finds itself operating in, and thus, journalists’ articulation of “power” and “freedom” informs the experiments with power/freedom I explore in the DWSN below.

I see two main threads in news reporting about the dark web, with one dominant and the other less dominant but quite prevalent. First, there is the conception of the dark web as entirely composed of illegal or taboo activities and in need of policing. Second, there is the idea that the dark web can preserve a valued liberal freedom: freedom of speech. Thus, what appears in this media ideology is a “reciprocal and incompatible” (Foucault, 2006: 529; Koopman, 2013: 163) relationship between power and freedom.

The popular media coverage of the dark web is redolent of moral panics that have been associated with Internet culture over the past 35 years, such as the panic about computer hackers and phone phreaks in the 1980s (resulting in the arrest of many young computer users); the US Congress’ Communications Decency Act of 1996, brought on by moral panic about pornography on the Web; and the US Congress Deleting Online Predators Act of 2006, inspired by moral panics over pedophiles on Myspace. Especially since the media coverage of the Silk Road drug market bust and the Freedom Hosting child pornography (CP) bust, both in 2013 (Borland, 2013), the dark web is currently inspiring similar panics centering on fears of CP, the drug and gun trade, and killers for hire.

One Sun headline is a series of adjectives that bind the dark web with CP: “Child sex dark web targeted” (Wooding, 2013). A reporter for The Age notes the dark web is “where pornographic images to satisfy the most depraved tastes can be downloaded” (Ormsby, 2012). The dark web is “a hub for black markets that sell or distribute drugs” (Pagliery, 2014). Gizmodo profiled the dark web gun store The Armory, asking “Could a band of anonymous weapon mongers prepare me and 19 imaginary compatriots for illegal warfare? If you’ve got a spare million or so, it looks like the answer is yes” (Biddle, 2012). The Daily Mail calls Tor a “seething matrix of encrypted websites” where one could hire hitmen for US$10,000. “So for those looking to bump off a difficult acquaintance, all they have to do is enter the Deep Web—known also as the ‘Dark Web’ or the ‘Undernet’—and search ‘hitman for hire’” (Mail Online, 2013). Implicit throughout this coverage is a call for more policing of the dark web (e.g. Biddle, 2012; Bingham, 2013; Gillespie, 2013; Henry, 2013; Murad and Hines, 2012).

Despite this dominant idea of the dark web as only useful to pedophiles, assassins, and junkies, recently more Internet users have started to use Tor and even hidden .onion sites (see Figure 1). This rise has been attributed in part to Edward Snowden’s revelations about the National Security Agency (NSA) and General Communication Headquarters (GCHQ) surveillance of the Internet (Borland, 2013). For privacy purposes, this is a positive development; more Tor users means more traffic and thus more obfuscation of user identities (Dingledine et al., 2004). This ties in with a secondary thread in popular coverage of the dark web: its affordances for journalists, activists, and whistleblowers who want to speak freely, despite state monitoring of the Internet. Many of the above-cited stories note that anonymizing software such as Tor can benefit anyone who wants to dissociate speech from identity, including political dissidents (e.g. Pagliery, 2014). As National Public Radio reported,

Tor’s executive director is working with victims of domestic abuse who need to communicate without being tracked by their abusers. Tor is also used by Chinese dissidents who can’t access sites like Twitter. And it became a valuable tool during the Arab Spring. (Rath, 2014)

Conclusion: DWSN as an experiment with power/freedom

What do we make of these power/freedom tensions on the DWSN? As Foucault argues in “What is Enlightenment?,” critique “is a historical investigation into the events that have led us to constitute ourselves and to recognize ourselves as subjects of what we are doing, thinking, saying” (Foucault, 1984: 45). Moreover, he argues, such a critique must be about experimenting with such historically imposed limitations. The members and administrators of the DWSN are doing this critical work—tracing, deploying, and working against both the historical limits of mainstream social networking and the historical limits of the media ideology of the dark web. They know the dark web is seen in popular culture as only used by drug traders and child pornographers. They also know that mainstream social media—a “safe” space on the clear web—presents the problems of surveillance and the commodification of personal data. To experiment with both historical conditions, the DWSN is reverse engineering (Gehl, 2014) the functionality and cultural practices of sites such as Facebook and Twitter and porting these functions onto the dark web.

The DWSN is building a dark web space that is productive in two senses. First is the “be the media” sense of mainstream social media sites: the now-classic narrative about social media is that the user is in control and that all of us are “producers” making our own culture through digital creativity (e.g. Bruns, 2008). However, the DWSN is also bringing software- and culturally developed social media restrictions to the dark web. In other words, it is porting in the standards of social media sites such as Facebook and Twitter: centralization, surveillance, prohibitions on actions, and the channeling user activity to algorithmic and interface-driven ends. Both of these interact in the heterogeneous assemblage, that is, the DWSN.

This brings me to DWSN’s specific assemblage of power/freedom. I am mindful of Foucault’s (2008: 186) call to specify and concretize categories such as power and freedom, echoed by Coleman’s (2010) call to “localize” and specify digital media (p. 489). Both Foucault and Coleman recommend empirical methods (i.e. genealogy, ethnography) to trace contexts and assemblages and avoid narrowing of any object qua universalizing it (see also Koopman, 2013).

In this vein, I would offer that the specific, concrete power/freedom practices of the DWSN center on practices of anonymity and infrastructure. Although at first glance anonymity might simply appear to be a form of freedom and infrastructure simply a form of power, they are reciprocal and incompatible in that each contains elements of each. That is, there are moments we can speak of “anonymity-power”: for example, the disciplinary admonitions to be anonymous on the DWSN, as well as the advantages of using a pseudonym to administer a site. We can also speak of “infrastructural-freedom”: the mix of the infrastructure of existing .onion hidden services as well as open-source social media software packages to allow for a complex form of dark web communication and social expression that cannot exist on the “clear web.” Moreover, this form of infrastructure-freedom is available to all who learn how to navigate it; it is a freedom that comes from the technical skill needed to find the DWSN.

We can of course speak of anonymity-freedom: the use of anonymous communication to explore ideas that are marginalized in more mainstream contexts, including proverbial “third rail” issues like suicide, violent political change, or pedophilia. As one DWSN member told me, if “I want to talk about things illegal in my country or report some abuse I can without fear of retaliation. This is it. Nothing exceptional.” And we can speak of infrastructure-power: after all, the administrators can delete posts or accounts with relative impunity due to the centralization of the DWSN (I have witnessed this many times).

I see the DWSN as an experiment with power and freedom through anonymity and infrastructure, an experiment of going beyond historically imposed limits. In terms of social networking, the anonymous elements of the DWSN are a far cry from what we now would recognize as mainstream social media, which involves real-world identities increasingly linked to consumer preferences and nonpolitical control (Gehl, 2013). Rather, the DWSN is not-for-profit and thus is not interested in producing its members as niche audiences to be sold to marketers. Moreover, the DWSN appears to be—although I cannot ever say for certain—virulently dedicated to protecting members against law enforcement and state power, something that cannot be said of Facebook, which after all has patented a means to hand user data over to governments (United States Patent, 2013).

Additionally, the DWSN is also dedicated to fighting the dominant conception of the dark web as a place that only the vilest among us want to be. Rather than fleeing the ever-supervised, ever-controlled Internet to some hidden, carefree corner where total anonymous freedom takes control, the DWSN shapes internal discourses in order to develop itself as (as its About page puts it) “a safe and moderated environment for the productive exchange of information.” This is not a free-for-all, but neither is it space where everything is controlled and thus happy (as Facebook seemingly wants to be).

The point is power and freedom always operate on one another. To challenge surveillance power as found incorporate social media and in state surveillance agencies and to challenge the ideology of an anything-goes dark web, the DWSN deploys an assemblage of anonymity and infrastructure to create an experimental “space-time” beyond the historical conditions it finds itself in (Deleuze, 1995: 172). Power and freedom “must be deployed simultaneously so that we can work within the internal tensions of their relationships” (Koopman, 2013: 169). To transform one, we have to transform the other in experiments. Such “experimental freedom perhaps does not make for good cinema on the blockbuster model. But it does make and may make further, for good practices of freedom” (Koopman, 2013: 174). The quiet, hidden, clear web-leery DWSN is just such an experiment, one that its members and administrators are always tinkering with—sometimes well, sometimes poorly, and never with guarantees. As Wendy Chun (2006) argues in Control and Freedom, “From our position of vulnerability, we must seize freedom that always moves beyond our control, that carries with it no guarantees but rather constantly engenders decisions to be made and actions to perform” (p. 30).

This active curation/moderation process is coming off an increasingly routine (non-linear or deterministic) cycle of content production moderation-displacement and percolation (see Figure 1). In this cycle, surface Web platforms often act as an initial host for potentially false, hateful, or otherwise malicious content. The platform hosting the content can then move to limit the information through a variety of strategies such as delisting, demonetization, account bans, and de-hosting, among others. Often, while this content might migrate via so-called “hate highways” to other surface Web platforms (N.F. Johnson, et al., 2019) or in the short run be simply reposted on the same site, much of the most heavily moderated content eventually moves to the Dark Web in a process of displacement. Lastly, many of the ideas preserved and fostered in this distributed and anonymous environment then percolate back up to the surface Web, potentially starting the cycle anew. Small vignettes can illustrate the stages of this process.

In short, from an informational perspective, the surface Web and the Dark Web exist as an interconnected whole. Content moderation efforts on the surface Web are at best partial, with the Dark Web providing an alternative core from which malicious content producers and consumers can continue their routine. Working to correct for society’s growing information problem requires attention to the whole system, not just the happenings on the major surface Web hubs.

Notes

1.The “dark web” is also often referred to as the “deep web,” but journalists and technologists who discuss the dark web take pains to distinguish the two. The “deep web” is defined as all the material online that commercial search engines such as Google and Bing cannot access (Bergman, 2001). This includes databases, the information behind login barriers, and automatically generated content that only appears online for brief periods. This content is not indexed by search engines, but it still can be reached with a standard browser (e.g. a stock installation of Firefox). This is distinct from the “dark web,” which requires special software to access.

2. The specific form of power that journalists call for is not quite clear. It could be what Foucault called discipline, what Deleuze called control, what Lazzarato or Neidich called nonpower, or what Galloway called the protocol. Or it could take another shape, perhaps a mix of other forms across multiple national and local contexts. Tracing the specific contours of the assemblage of surveillance, manipulation, and punishment that would be needed to “clean up” the dark web is the subject of another study.

3. Although the administrators and members of the DWSN are wary of popular coverage of their social network, those that I spoke with were welcoming of my pursuing an academic publication. But because of their concern about “clear web” coverage, I ultimately decided to keep the real name of the site redacted.