Top sites (and maybe the NSA) track users with “device fingerprinting” (original) (raw)
May make it easier to follow privacy-minded users on the darknet.
Close to 1.5 percent of the Internet's top websites track users without their knowledge or consent, even when visitors have enabled their browser's Do Not Track option, according to an academic research paper that raises new questions and concerns about online privacy.
The research, by a team of scientists in Europe, is among the first to expose the real-world practice of "device fingerprinting," a process that collects the screen size, list of available fonts, software versions, and other properties of the visitor's computer or smartphone to create a profile that is often unique to that machine. The researchers scanned select pages of the top 10,000 websites as ranked by Alexa and found that 145 of them deployed code based on Adobe's Flash Player that fingerprinted users surreptitiously. When they expanded their survey to the top one million sites, they found 404 that used JavaScript-based fingerprinting. The researchers said the figures should be taken as the lower bounds since their crawlers weren't able to access pages behind CAPTHCAs and other types of Web forms. Mainstream awareness of fingerprinting first surfaced three years ago following the release of research from the Electronic Frontier Foundation.
Device fingerprinting serves many legitimate purposes, including mitigating the impact of denial-of-service attacks, preventing fraud, protecting against account hijacking, and curbing content scraping, bots, and other automated nuisances. But fingerprinting also has a darker side. For one, few websites that include fingerprinting code in their pages disclose the practice in their terms of service. For another, marketing companies advertise their ability to use fingerprinting to identify user behavior across websites and devices. That suggests device fingerprinting may be used much the way tracking cookies are used to follow people as they browse from site to site, even though fingerprinting isn't covered by most laws governing cookies and websites' Do Not Track policies. And unlike user profiling that relies on "stateful" browser cookies that are usually easy to delete from hard drives, most end users have no idea that their computers are being fingerprinted, and they have few recourses to prevent the practice.
"Device fingerprinting raises serious privacy concerns for everyday users," the researchers wrote in a recently published paper. "Its stateless nature makes it hard to detect (no cookies to inspect and delete) and even harder to opt-out. Moreover, fingerprinting works just as well in the 'private-mode' of modern browsers, which cookie-conscious users may be utilizing to perform privacy-sensitive operations."
More troubling, device fingerprinting may have given the National Security Agency and its counterparts around the world an avenue to identify people using the Tor privacy service. As disclosed in an installment of previously secret NSA documents published last week by The Guardian, the spy agency is capable of injecting script redirections into the traffic of Tor users. Slide 16 of an NSA presentation titled Tor Stinks included the excerpt: "Goal: ... Ignore user-agents from Torbutton or Improve browser fingerprinting? Using javascript instead of Flash?"
The Firefox browser that ships with the Tor Browser Bundle has long attempted to prevent fingerprinting by limiting the customizable properties that are available to users. It also placed a cap on the number of fonts a webpage can request or load. The fingerprinting researchers found a way to bypass the font cap by making use of the Web programming property known as CSS font face. The researchers reported their findings to Tor developers, who have since patched the weaknesses.
Orbitz, T-Mobile, Western Union, and Pokerstrategy among the players
The researchers said their lawyers advised them not to provide an exhaustive list of the 404 or more websites that hosted tracking code. Responding to questions from Ars, researcher Gunes Acar of KU Leuven University in Belgium said that they included orbitz.com, tmobile.co.uk, pokerstrategy.com, anonymizer.com, westernunion.com, and t-online.de. He stressed that his team may have missed some sites given the limitations of their scanning technology. Tracking code based on Adobe Flash is particularly time consuming to detect because it must be decompiled and manually analyzed. As a result, the researchers scanned only 10,000 sites and limited the searches to homepages, even though some sites are known to deploy tracking code on registration pages and other subsections.
Two of the few websites named in the paper are privacytool.org and anonymizer.com, both owned by a company called Anonymizer Inc. The former site offers a Java applet that allows visitors to test how easy they are to track online. Before users can run it, the site discloses what information will be gathered and warns, "Data obtained from the browser like lists of plug-ins or fonts can be used to identify your computer." Anonymizer.com, in sharp contrast, ran largely the same fingerprinting scripts on its homepage without making any mention that it was compiling a list of fonts and plug-ins that could be used to identify an end user's computer.
"Finally, note that while privacytool.org offers informed choice to its users, who may voluntarily execute the script, the fingerprinting scripts that run in the anonymizer.com homepage are invisible to users and run by default," the researchers reported.
Anonymizer.com representatives didn't respond to messages seeking comment for this article.
Going mainstream
The researchers said many website operators may have no idea that fingerprinting is taking place on the pages they maintain. In one case, a font-probing script was embedded into a button users clicked to donate bitcoins to the site owner. The researchers uncovered 15 third-party providers of fingerprinting services, with names including BlueCava, Perferencement, CoinBase, and MaxMind.
"What surprised us was the variety of the companies providing this service," Acar wrote in an e-mail to Ars. "We knew it was used to secure authentication for online banking or retail, but what we found out is it's becoming mainstream. We even found that fingerprinting scripts are embedded in widgets or ad banners."
He said end users who want to prevent their systems from being fingerprinted have few options other than to run Tor. Disabling JavaScript and Flash in the browser reduces some of the information websites can collect, but it does nothing to stop font probing, including the attack on older versions of the Tor Browser. What's more, disabling both Flash and JavaScript ironically creates a configuration that rare. There are a variety of services that test test how identifiable a specific browser is, including this one from the EFF.
The paper—titled "FPDetective: Dusting the Web for Fingerprints"—will be presented next month at the 20th ACM Conference on Computer and Communications Security in Berlin. The research is important because it highlights a practice that few people have any idea is taking place. Readers shouldn't view it as an indictment of fingerprinting itself or the websites that engage in it. But it's fair to read the findings as clear evidence that fingerprinting probably won't remain a nascent or niche practice for much longer. Websites that want visitors' trust should disclose how and when fingerprinting is used and take steps to limit its harm to user privacy.
Dan Goodin is Senior Security Editor at Ars Technica, where he oversees coverage of malware, computer espionage, botnets, hardware hacking, encryption, and passwords. In his spare time, he enjoys gardening, cooking, and following the independent music scene. Dan is based in San Francisco. Follow him at @dangoodin on Mastodon. Contact him on Signal at DanArs.82.
