BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Private Statistics and Their Applications to Distributed Learning:
  Tools and Challenges - Dr Emiliano De Cristofaro - University College Lon
 don
DTSTART:20180207T161500Z
DTEND:20180207T171500Z
UID:TALK98440@talks.cam.ac.uk
CONTACT:David Greaves
DESCRIPTION:Large-scale distributed collection of contextual information i
 s often essential in order to gather statistics and train machine learning
  models. The ability to do so in a privacy-preserving way enables a number
  of computational scenarios that would be hard\, or outright impossible\, 
 to realize without strong security guarantees. In this talk\, we present t
 he design and deployment of practical techniques for privately gathering s
 tatistics from large data streams. We build on efficient cryptographic pro
 tocols for private aggregation and on data structures for succinct data re
 presentation\, namely\, Count-Min Sketch and Count Sketch. We then show ho
 w to use these techniques to instantiate real-world privacy-friendly syste
 ms\, supporting\, among others\, recommendations for media streaming servi
 ces and crowd-sourced mobility analytics.\n\nWe then focus on how to ident
 ify and quantify possible privacy leakage from the aggregate statistics. W
 e frame the problem in terms of the advantage an adversary has\, from the 
 aggregates\, in profiling or inferring membership target users\, and prese
 nt two novel frameworks to quantify such leakage vis-a-vis inference attac
 ks even when differential privacy protections are used.
LOCATION:Lecture Theatre 1\, Computer Laboratory
END:VEVENT
END:VCALENDAR