CS Colloquium on Wednesday, June 9, 2004

Title: Private Transaction Management through
       Autonomous Oblivious Data Mining

Speaker: Aggelos Kiayias
         University of Connecticut

Date: June 9, 2004

Time: 11am

Place: MCS 135

Abstract:

In many systems, privacy regulations/concerns mandate privacy (anonymity)
and personal data protection provisions. On the other hand, system
management and system integrity constraints may require monitoring user
activities and aggregate activity reporting (for billing, statistics,
misbehavior detection, etc.).  In this work, we suggest a new concept
which we call "private transaction management" that achieves the above
conflicting goals simultaneously. It allows users to transact anonymously,
yet the layer of system management is capable of calculating and reporting
the transaction data at an aggregate level. Furthermore, the aggregate
information may reveal unacceptable patterns of user's misbehavior, in
which case the system may react by privacy revocation.  To realize the
concept, we design a new credential mechanism for private transactions
which allows the above balanced functionality and combines ideas from
group signatures with a novel component called "autonomous oblivious data
mining engine." This component enables efficient server-based evaluation
of aggregate functionalities from the raw transactions data while
maintaining user privacy.

[Joint work with Shouhuai Xu (U. Texas San Antonio) and Moti Yung (Columbia U.) ]

Host: Leonid Reyzin (http://www.cs.bu.edu/~reyzin)