Title:	Managing Data Streams in Advertising Networks
Date:	November 11, 2005 (Friday)
Time:	2:30 p.m. - 3:30 p.m.
Venue:	Room 121, 1/F, Ho Sin-hang Engineering Building, The Chinese University of Hong Kong, Shatin, N.T.
Speaker:	Prof. Amr El Abbadi Department of Computer Science University of California at Santa Barbara (In collaboration with Ahmed Metwally and Divyakant Agrawal)

ABSTRACT:

This research is primarily motivated by data streams in the setting of Internet advertising commissioners, who represent the middlemen between Internet publishers (entities which operate sites/pages on the internet), and Internet advertisers (entities which like to place advertisements on pages of Internet publishers). Since publishers earn revenue on the traffic they drive to the advertisers' Websites, there is an incentive for them to fradulently increase the number of clicks their sites generate. This phenomenon is referred to as hit inflation. One of the advertising commissioner's roles is to detect fraud that takes place on their advertising network. Another objective of the advertising commissioner is to optimize the nature of advertisements displayed to different surfers. In this talk, we develop an integrated approach for solving both problems of finding the most popular k elements, and finding frequent elements in a data stream. This approach can efficiently allow the advertising commissioners to dynamically change the advertisements displayed on the publishers' pages based on the customers recent behaviour. It also forms the basis for solving the problem of hit inflation. One of our project goals is to identify ways publishers use to falsify traffic. Publishers use various ways to simulate traffic, and there is no one unified way that can detect all the fraud techniques. We have devised efficient solutions that can detect some hit inflation techniques online by analyzing the click streams on the fly. We will discuss the most difficult-to-detect known technique employed by publishers. This technique involves a coalition of one or more Websites with the fraudulent publisher. This problem motivates a novel technique that detects such fraud by finding associations between websites in a stream of HTTP requests.

BIOGRAPHY:

Amr El Abbadi received his Ph.D. in Computer Science from Cornell University. In August 1987 he joined the Department of Computer Science at the University of California, Santa Barbara, where he is currently a Professor. In 1990 he was a visiting professor at the University of Campinas in Brazil, in 1994 a visiting scientist at IBM Almaden Research Center, in August 1998 a visiting lecturer at the Swedish Institute of Computer Science in Stockholm, Sweden and in July 1999 a visiting researcher at IRISA at the University of Rennes in France. He has served as area editor for Information Systems: An International Journal, an editor of Information Processing Letters (IPL) and Associate editor of the Bulleten of the Technical Committee on Data Engineering. He was Vice Chair of the 1999 International Conference on Distributed Computing Systems, Vice Chair for the International Conference on Data Engineering 2002, group leader for the International Conference on Management of Data (SIGMOD) 2005, and the Americas Program Chair for the 2000 International Conference on Very Large Data Bases (VLDB). Currently, he is a board member of the VLDB Endowment. Prof. EL Abbadi's main research interests and accomplishments have been in understanding and developing basic mechanisms for supporting distributed information management systems, including databases, digital libraries, peer-to-peer systems, and spatial databases.

Enquiries: Miss Temmy So at tel 2609 8444

For more information, please refer to http://www.cse.cuhk.edu.hk/seminar