University researchers have developed a technique that governments and Internet service providers could use to bypass secured Internet connections and gather valuable personal information.
The "analysis attack" on HTTPS traffic had an 89 percent accuracy rate in determining the Web pages a person visited, according to University of California, Berkeley, researchers. Such tracking made it possible for the researchers to gather information on medical conditions, sexual orientation, financial status and whether a person is involved in a divorce or bankruptcy proceeding.
The study looked at more than 463,000 page loads on 10 widely used, industry-leading websites. Healthcare sites included those of the Mayo Clinic, Planned Parenthood and Kaiser Permanente; financial sites belonged to Wells Fargo, Bank of America and Vanguard; legal services sites belonged to the American Civil Liberties Union and Legal Zoom; and video-streaming sites included Netflix and YouTube.
For the attack to work, snoops would have to be able to visit the same Web pages as the victim, which would enable the attackers to identify packet patterns in encrypted traffic that would be indicative of different Web pages.
"It would be like if somebody gave you a bicycle but took it apart and wrapped each piece individually," Brad Miller, co-author of the study told CSOonline Monday. "You would quickly notice that there were two big packages which look like wheels, a frame, a squiggly one that corresponds to a chain, etc.
"It's the same way with a Web page. Because we watch each of the parts be delivered individually, there ends up being so much information which you can observe without decrypting the packets that you can quite likely figure out the exact Web page."
The attackers must also be able to observe victim traffic, which would allow them to match those packet patterns with the ones going to particular Web pages.
The researchers also developed a defense that involved reducing the amount of packet information an attacker could gather. The technique lowered the accuracy of identifying Web pages visited by people from 89 percent to 27 percent.
The research has important privacy implications. Being able to examine user activity on a healthcare site could reveal medical conditions, which could lead to discrimination or could be sold to advertisers looking to pitch products.
Monitoring legal site traffic could uncover a divorce, bankruptcy or immigration status, while analyzing traffic on a banking site could provide insight on whether a person has children, is in a long-term relationship or is in a high-income bracket.
Any company with access to HTTPS traffic, such as Internet service providers and commercial chains of Wi-Fi access points, could gather data on users despite the encryption and sell the information to advertisers, the study said.
Sign up for Computerworld eNewsletters.