Workshop 9 PrivacyEnhanced Personalization

1
Workshop 9Privacy-Enhanced Personalization

• Workshop co-chairsAlfred Kobsa Lorrie Cranor
• http//www.isr.uci.edu/pep05/

2
Program committee

• Alfred Kobsa, University of California, Irvine,
  CA
• Lorrie Cranor, Carnegie Mellon University,
  Pittsburgh, PA
• John Canny, University of California, Berkeley,
  CA
• Clare-Marie Karat, IBM Watson Research Ctr., 
  Hawthorne, NY
• Judy Kay, University of Sydney, Australia
• Sarah Spiekermann, Humboldt University, Berlin,
  Germany
• Loren Terveen, University of Minnesota,
  Minneapolis, MN

3
PEP Program

• Full-day workshop
• 16 submissions received
• 1 position paper
• 1 invited industry talk
• Caspar Bowden, Microsoft Chief Privacy Advisor
  for Europe, Middle East and Asia
• The Need for an Identity Meta-System
• 4 long papers
• 5 short papers
• Discussion

4
Position paper

• Chris Demchak proposed a framework for escrowed
  identity, designed to protect privacy but allow
  for identification of bad actors

5
User studies

• Sara Spiekermann discussed privacy issues with
  RFID tags and a way of evaluating how users feel
  about proposed privacy enhancements
• Evelien Perik discussed a study to see whether
  user privacy preferences varied depending on type
  of data in a music recommendation system
• Thea van der Geest discussed a project to address
  privacy concerns associated with an e-government
  web site that will use profiling to reduce the
  amount of forms people have to fill out

6
Catering to privacy requirements

• Yang Wang discussed a software product line
  approach to make it easier to automate the
  handling of privacy constraints in web
  personalization based on the needs of specific
  users/jurisdictions
• Judy Kay presented a framework for privacy in a
  ubiquitous personalized environment

7
Selective access to user data

• Jim Greer discussed a design for a single sign-on
  identity management system that avoids most of
  the pitfalls of systems such as Microsoft
  Passport and Liberty
• Elke Franz discussed a personalized tutoring
  system and various interfaces to allow users to
  specify when the system should record user data
  as part of different contexts

8
Privacy in recommender systems

• Schlomo Berkovsky discussed an approach to
  increasing privacy in collaborative filtering by
  allowing users to store their own profiles and
  release obfuscated versions of them in response
  to a recommendation request
• John Riedl discussed the issue of shilling in
  recommender systems and raised a number of
  questions about privacy in recommender systems

9
Discussion - A few highlights

• Much of this work is in early stages, many of us
  have more questions than answers
• Identified profile management as problem that
  must be addressed to improve privacy in
  personalization
• Does less information mean more privacy?
  Sometimes which information you keep is critical
• Important for user modeling community to think
  more about privacy, also collaboration needed
  with security/policy/UI communities
• Interest in organizing similar workshop at
  another conference
• Interest in journal special issue that would
  reach interdisciplinary audience