Ph.D. Thesis Defense Announcement

Title: MITIGATING INFORMATION MANIPULATION

Xinyu Xing

Ph.D. Candidate

School of Computer Science

College of Computing

Georgia Institute of Technology

Date: August 17, 2015 (Monday)

Time: 12:00 PM - 2:00 PM (ET)

Location: Klaus 3126

Committee:

------------

Dr. Wenke Lee (Advisor, School of Computer Science, Georgia Tech)

Dr. Nick Feamster (Co-advisor, Department of Computer Science, Princeton University)

Dr. Mustaque Ahamad (School of Computer Science, Georgia Tech)

Dr. Michael Bailey (Department of Electrical and Computer Engineering , University of Illinois at Urbana-Champaign )

Dr. Hongyuan Zha (School of Computational Science and Engineering, Georgia Tech)

Abstract:

The advent of information services introduces many advantages, for example, in trade, production and services. While making important descisons today, people increasingly rely on the information gleaned from such services. Presumably, as such, information from these services has become a target of manipulation.  

During the past decade, we have already observed many forms of information manipulation that misrepresents or alters reality. Some popular manipulation -- we have ever witnessed on the Internet -- include using black hat SEO techniques to drive up the ranking of a disreputable business, creating disinformative campaigns to conceal political dissidence, and employing less-than-honest product assessments to paint a rosy picture for inferior wares.  Today, emerging web services and technologies greatly facilitated and enhanced people's lives. However, these innovations also enrich the arsenal of manipulators.  

The sheer amount of online information available today can threaten to overwhelm any user. To help ensure that users do not drown in the flood of information, modern web services are increasing relying upon personalization to improve the quality of their customers' experience. At the same time, personalization also represents new ammunition for all manipulators seeking to steer user eyeballs, regardless of their intents. In this thesis, I demonstrate a new unforeseen manipulation that exploits the mechanisms and algorithms underlying personalization. To undermine the effect of such  manipulation, this thesis also introduces two effective, efficient mitigation strategies that can be applied to a number of personalization services. 

In addition to aforementioned personalization, increasingly prevalent browser extensions augment the ability to distort online information. In this thesis, I unveil an overlooked but widespread manipulation phenomenon in which miscreants abuse the privilege of browser extensions to tamper with the online advertisement presented to users. Considering that online advertising business is one of the primary approaches used to monetize free online services and applications available to users, and reckless ad manipulation may significantly roil advertising ecosystem, this thesis scrutinizes the potential effect of ad manipulation, and develops a technical approach to detect those browser extensions that falsify the ads presented to end users. 

Although the thesis merely discusses several manipulation examples in the context of the Internet, the findings and technologies presented in this thesis introduce broad impacts. First, my research findings raise Internet users' awareness about pervasive information manipulation. Second, the proposed technologies help users alleviate the pernicious effects of existing information manipulation. Finally, accompanying the findings and technologies is publicly available open-source software and tools that will help an increasing number of users battle against the growing threat of information manipulation.