For a week in March, Website security firm CloudFlare analyzed the traffic hitting its customers’ sites from the anonymous Tor network.
The results of the study illustrate the double-edged nature of online anonymity. The Tor network—a peer-to-peer collection of volunteered servers linked together to create an anonymizing Web service—allows people in repressive countries to surf the Internet, enables activists to communicate freely and helps journalists evade government surveillance.
Yet, it also allows criminals to act with little fear of repercussions, because circumventing the anonymity provided by Tor is difficult, albeit not impossible.
In this case, the company found that nearly 94 percent of requests to CloudFlare customers’ Web sites coming from the Tor network were automated and malicious—with comment spam, vulnerability scanning, advertising click-fraud, content scraping and brute-force log-in attempts topping the list of attacks coming from the network.
Because these types of attacks produce a lot of requests, a small number of attackers can use automation to create a large footprint, Matthew Prince, co-founder and CEO of CloudFlare, told eWEEK.
“It doesn’t mean that 94 percent of users are bad, or that 94 percent of Tor traffic is bad,” he said. “It is a very small universe of bad actors that is causing this large problem for our customers.”
The Tor Project, the nonprofit foundation that maintains and develops Tor client and server software, criticized CloudFlare’s approach to Web security. Tor users have complained that sites that use CloudFlare often throw up CAPTCHAs—automated tests to detect whether a visitor is a human or a bot—that can block people from reaching those sites, Mike Perry, Tor browser and Tor performance developer, stated in a blog post responding to the CloudFlare data.
“We suspect this figure is based on a flawed methodology by which CloudFlare labels all traffic from an IP address that has ever sent spam as ‘malicious,’ ” he wrote. “Tor IP addresses are conduits for millions of people who are then blocked from reaching websites under CloudFlare’s system.”
CloudFlare’s Prince rebutted Perry’s assertion, but offered only general points about how the company measures bots online, saying that the company uses a variety of techniques to determine whether a request is from an automated source, including creating content that is visible only to bots and turning off CloudFlare protections on certain Internet servers to use as controls.
The CloudFlare study is unlikely to settle the debate over the relative benefits and drawbacks of the Tor network. The network, created in the ’90s by a group of researchers and computer scientists affiliated with the U.S. Department of Defense, was originally conceived as a way to hide the Internet addresses of U.S. intelligence gatherers, so that their systems would not automatically be blocked. Because hiding in a crowd of users works only if there is a crowd of users, the U.S. government—through grants from the National Science Foundation—continues to support the project. Currently, there are about 2 million daily Tor users.
The benefits of the Tor Network rarely garner media attention, however. Instead, coverage is dominated by new reports of the more seedy side of the anonymity network, the so-called Dark Net. These include the takedown of online bazaars for drugs and illegal goods—such as the Silk Road markets, the shuttering of child porn rings using Tor, and the use of the network for malware communications.
Traffic Studies Reveal Complex Picture of Tor’s Role on ‘Dark Web’
The focus on Tor’s bad side has resulted in people wanting to shut down the Dark Net and, by extension, the Tor network, Eric Jardine, research fellow for the Centre for International Governance Innovation (CIGI), told eWEEK. More than 70 percent of people across the globe want to shutter the Dark Net, CIGI found in its annual 2016 CIGI-Ipsos Global Survey on Internet Security and Trust, which polled more than 24,000 Internet users.
“A lot of people really don’t know what the Tor network is as a technology and what its various functions are,” he said. “And they see a news story—such as a child-abuse ring or an illegal marketplace—and they have a knee-jerk reaction, saying, ‘We don’t need this. Shut it down.'”
Jardine argues that the Tor Network has a marketing problem. A simple name change—one online user proposed the “Freedom Network”—might go a long way toward changing opinions, he said. Considering that CIGI also found that only 38 percent of people trust that their activities on the Internet are not monitored, a continued focus on privacy should help as well.
Looking at the data is unlikely to answer the question of whether Tor is a haven for bad actors or for people trying to fight oppression overseas. While CloudFlare’s study arguably shows that a small number of automated systems can abuse the network to create a large number of attacks on Web sites, other studies have found different results.
In its own look at Tor traffic, Akamai found that only 0.3 percent of requests coming from a Tor exit node attacked Web sites. Yet, Akamai focused on a narrow definition of attacks—requests that attempted to exploit a Web application, such as SQL injection, cross-site scripting and command injection. Such attacks tend to be much more focused and produce less bandwidth than the attacks viewed by CloudFlare.
Akamai also, however, found that requests from Tor exit nodes had an equal likelihood as non-Tor traffic to conduct a legitimate commercial transaction, suggesting that Tor users may be just as valuable to business sites as non-Tor visitors are.
“They are not just there to surf the Internet, but to shop the sites,” Larry Cashdollar, an aptly named senior security response engineer for Akamai, told eWEEK.
In a separate study, bot-blocker Distil Networks—whose data may be more comparable to CloudFlare’s—found that 48 percent of traffic from Tor and other proxies violated its rules for legitimate traffic. A small number of users can easily create a large volume of malicious traffic, Rami Essaid, CEO of Distil, told eWEEK.
“You can have a handful of bad actors that can pollute the Tor IPs, since the fundamental premise of Tor is to not assign a static IP [to] an individual,” he said.
CloudFlare’s Prince recognizes that the company’s study looks at a segment of Tor traffic that applies most to its customers. He stressed that the company’s classification of requests from Tor likely represents only a minority of traffic going through the anonymizing network.
In addition to all the traffic headed to Websites that does not consist of CloudFlare customers, about 60 percent of traffic on the Tor Network is peer-to-peer file-sharing, which is never seen by CloudFlare, he said. Furthermore, malware that uses Tor for command-and-control traffic would also not be visible to CloudFlare.
“So inherently, we see only the sliver of Tor traffic that goes to HTTP and HTTPS sites,” Prince said.