Apple Adding Bad Data to Give Good Privacy

Apple is making widespread use of 'differential privacy' to protect individual identities while deriving useful insights from statistics. The technique involves adding incorrect 'noise' to big data, making it far more difficult to identify those supplying it.

Interviewed for www.appleinsider.com , University of Pennsylvania mathematician Prof. Aaron Roth, a pioneer of the technique and now an advisor to the tech giant, said Apple was the first company to make it 'a central focus of its data collection efforts' - although Google has used it with the Chrome browser for its Google Report Project previously.

Differential privacy is said to provide an effective means of obscuring individual answers while still getting representative samples. The first academic paper on the method was issued in 2006, and in 2014 Roth published the book 'The Algorithmic Foundations of Differential Privacy', with Cynthia Dwork.

Apple will roll out the approach in the autumn, when it releases an update to its iOS for iPhone and iPad, using it in analysis of search behaviour, and use of applications, web sites, emojis and new slang expressions'. According to a report quoted in the Wall Street Journal, it will allow Apple engineers to spot usage patterns across large numbers of users without violating an individual's privacy.