Randomisation – IBM's answer to Web privacy
Have your cake and eat it
IBM Corp's new Privacy Institute has decided that randomization may be the key to protecting consumer privacy on the web while also providing e-businesses with informative metrics on their customers.
Thursday last week, the company said it has developed software that ensures consumers' sensitive data never leaves their computers in an accurate form, but can be reassembled at the back end in aggregate. IBM is looking for partners to develop the software.
"What we wanted to do is to protect users' privacy... but we also wanted businesses to get good information," said Rakesh Agrawal, who headed up the project. "The dilemma was how to balance those two objectives, how do you have your cake and eat it?"
The answer Agrawal, along with Ramakrishnan Srikant, came up with is an algorithm for reconstructing previously randomized data sets in such a way that the margin of error is only between 5% and 10% with a randomization of 100%, IBM claims.
In an example, a consumer registering for a web site truthfully enters their age as 30. Software in the page, perhaps a Java applet, is set to randomly add or subject years in a range of, say, five years, before submitting the data to the site, so submits the user's age as 26.
This happens with all the other users of the site, until the company has a data set of a few thousand individuals (that's all that's needed to get an accurate picture, Agrawal says) but with wildly skewed data that does not represent the true demographics of the users.
That's when IBM's special sauce kicks in. Before allowing the data to be input to a data mining application, IBM's software "corrects" the randomized data to provide a "close approximation of the true distribution". How, exactly, IBM was not ready to disclose, but it involves knowing what the range of randomization was in the first place.
"The intention now is to look for people that want to go into partnership with a beta," said IBM spokesperson Kendra Collins. The software has obvious commercial potential, but whether it will solve the problem of people entering false information is another question
© ComputerWire. All rights reserved.