by JP Mangalindan
One feature Netflix users have always taken for granted, privacy, might be in danger if they're one of the 480,000 members whose information was shared as part of a contest to improve the company's popular movie recommendation system. According to Wired, a closeted lesbian mother going by the pseudonym "Jane Doe" filed a lawsuit yesterday against the rental giant, claiming it had, among other things, broken a federal privacy law protecting citizens' video rental information.
The backstory: Netflix started a contest back in September 2006 that gave 50,000 contestants vying for $1 million two sets of data. The first included 100 million movie ratings, the date of the rating, subscriber IDs, and movie info culled from some 480,000 customers. The goal? Create a movie recommendation algorithm at least 10-percent more accurate than Netflix's own.
Though user data had been anonymized through a common method called perturbation, that didn't stop two savvy University of Texas researchers from figuring out the real identities, political leanings, and even sexual orientation of several Netflix users, by cross-referencing data given by the company to reviews posted on IMDB. Which is why Doe is concerned: if information like her sexual orientation became available, it might negatively affect her and her family's lives, she stated in her complaint. But, we point out: the Netflix findings aren't public, and the ability to re-link identities to connect to other Web profiles is neither encouraged nor part of the contest. Netflix never gives users the option to state a sexual preference, so the claim that raw data can be used to predict sexuality outside of Netflix is either suggesting movie choice determines type of person, or it's Netflix's fault for allowing researchers to look outside of the given data set.
However, Doe does bring up a crucial point, as her suit also aims to stop Netflix from starting the second phase of its contest, which would provide even more information about customers, like ZIP codes, ages, and gender. If that comes to pass, Wired remarks that it's 87-percent possible to take someone's age, tastes, and ZIP code and hunt them down on the Internet. With these tangible, real-world (and identifiable values), Netflix is breaching user privacy (especially if those accounts opt to be hidden).
Therefore, it appears the Netflix contest really did succeed in what it set out to do, which was accurately glean our preferences and choices via movie lists. So while movie recommendations might be more accurate going forward, remember: every time you rate a movie, song, or TV show, bits of yourself are released into the cyber-sphere.
From Switch.com and Wired.com
Sunday, December 20, 2009
by JP Mangalindan