Cross-posted at the Tech4Humanity Lab blog at VT.
This week, my new article with Eric Jardine and Gareth Owenson, “The potential harms of the Tor anonymity network cluster disproportionately in free countries,” came out in PNAS. Using a global sample of Tor users, we show that a higher percentage of Tor clients in politically free countries go to hidden services sites than in less free countries. Past studies have shown that much of the traffic to hidden services sites goes to cryptomarkets and child abuse imagery sites. The corollary is that higher % of clients in repressive regimes use Tor to access the surface web.
Academics can be fussy about a lot of things. For example, the term, “the Dark Web,” is contested and, so, in the article, we use the more precise phrase, “the Tor anonymity network.”
We also go to great pains to acknowledge that not all hidden services sites are socially harmful and not everything on the surface web is beneficial. Well, duh. But you would have to be willfully naïve not to recognize that the preponderance of hidden services traffic goes to content that is bad for society. Speaking for myself, I’m not terribly concerned about adults buying marijuana on a cryptomarket, but I think it would be better if it were legal, regulated, and taxed. Sales of arms, malware, child abuse content, and hard drugs are all cause for serious concern.
Our results, then, show that more of Tor’s potential harms are in mostly affluent, free countries, while a disproportionate share of its benefits – accessing news and information largely free of surveillance – tend to be enjoyed by people living under repressive regimes.
But what countries are we talking about? Like many articles in the social sciences, our paper focuses on the larger pattern. But, here – because this is a blog and, on blogs, you can do what you feel like doing – I’d like to dig into the particulars of individual countries. The figure below shows the percentage of a country’s average daily clients that go to hidden services by the total number of clients (on a log axis) in the observed period (most of 2019). As you can see, there are a number of countries (“Basically No Tor”), most of which are small and poor, which very few Tor clients. The “Bad Stuff in Small Countries” category refers to a cluster of countries that are small, usually politically free, and have an unusually high percentage of Hidden Services. These countries, including New Zealand and Liechtenstein, tend have relatively high cryptomarket use among few total users. There are some slightly less tiny countries like Finland and Iceland that are also politically free and have very active cryptomarkets. Still, as small countries, each individually accounts for just a fraction of overall daily volume of Tor clients.
Unsurprisingly, some of highest total volume countries for Tor use are big, rich, free countries, like the US, UK, Germany, and the Netherlands. Among the highest volume countries, they all have somewhat higher percentages of clients visiting hidden services.
By contrast, the other group of super-high volume countries, including Iran, Russia, and China, which all have repressive regimes. In these countries, a greater share of the clients are using Tor to visit surface web sites.
In short, among the countries that account for most of Tor’s users, they are divided among a) free countries with a disproportionate share of clients visiting hidden services and b) repressive countries where the overwhelming majority of Tor clients head to the surface web.