BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Engineering Privacy for Small Groups - Graham Cormode (University 
 of Warwick)
DTSTART:20161110T153000Z
DTEND:20161110T163000Z
UID:TALK69044@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:Concern about how to collect sensitive user data without compr
 omising individual privacy is a major barrier to greater availability of d
 ata. The model of Local Differential Privacy has recently gained favor and
  enjoys widespread adoption for data gathering from browsers and apps. Dep
 loyed methods use Randomized Response\, which applies only when the user d
 ata is a single bit.&nbsp\; We study general mechanisms for data release w
 hich allow the release of statistics from small groups.  We formalize this
  by introducing a set of desirable properties that such mechanisms can obe
 y. Any combination of these can be satisfied by solving a linear program w
 hich minimizes a cost function. We also provide explicit constructions tha
 t are optimal for certain combinations of properties\, and show a closed f
 orm for their cost. In the end\, there are only three distinct optimal mec
 hanisms to choose between: one is the well-known (truncated) geometric mec
 hanism\; the second a novel mechanism that we introduce\, and the third is
  found as the solution to a particular LP. Through a set of experiments on
  real and synthetic data we determine which is preferable in practice\, fo
 r different combinations of data distributions and privacy parameters.  &n
 bsp\;  Joint work with Tejas Kulkarni and Divesh Srivastava
LOCATION:Seminar Room 2\, Newton Institute
END:VEVENT
END:VCALENDAR