Identifying the optimal di ff erential private mechanisms for di ff erent users ?

Content Provider	Semantic Scholar
Author	ElSalamouny, Ehab Chatzikokolakis, Konstantinos Palamidessi, Catuscia
Copyright Year	2012
Abstract	The notion of differential privacy has emerged in the area of statistical databases to provide a protection for the sensitive information about participants in these databases. Without concerning the privacy protection, participants’ sensitive information can be leaked easily to an attacker by performing selected queries on such databases. Differential privacy is satisfied using a ‘randomisation’ mechanism which provides the user with a ‘noisy’ answer for her query instead of the exact answer. The privacy is therefore achieved at the cost of reducing the accuracy (or ‘utility’) of the user’s query answer. A trend of research has recently directed to finding the ‘optimal’ differentially private mechanisms which provide a trade-off between privacy and utility. The main challenge is that an optimal mechanism for a user depends on both the database query and the user’s side information about possible query results, modelled as a ‘prior’ probability distribution over these results. In this work we describe, for a general query and privacy level, a randomisation mechanism which satisfies differential privacy and at the same time is optimal for a class of users having various priors. We present the properties of this mechanism in terms of utility and privacy, and also characterise the class of users for whom it is optimal.
File Format	PDF HTM / HTML
Alternate Webpage(s)	http://www.lix.polytechnique.fr/Labo/Ehab.Elsalamouny/papers/privacy-avap-2012.pdf
Language	English
Access Restriction	Open
Content Type	Text
Resource Type	Article