Declassification of Information with Complex Filter Functions
Kurt Stenzel, Kuzman Katkalov, Marian Borek, Wolfgang Reif · 2016
Many applications that handle private or confidential data release part of this data in a controlled manner through filter functions.However, it can be difficult to reason formally about exactly what or how much information is declassified.Often, anonymity is measured by reasoning about the equivalence classes of all inputs to the filter that map to the same output.An observer or attacker that sees the output of the filter then only knows that the secret input belongs to one of these classes, but not the exact input.We propose a technique suitable for complex filter functions together with a proof method, that additionally can provide meaningful guarantees.We illustrate the technique with a DistanceTracker app in a leaky and a non-leaky version.