CEC Theses and Dissertations

Date of Award

2010

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Information Systems (DISS)

Department

Graduate School of Computer and Information Sciences

Advisor

Maxine S. Cohen

Committee Member

Francisco J. Mitropoulos

Committee Member

Sumitra Mukherjee

Abstract

Apparently fraud is a growth industry. The monetary losses from Internet fraud have increased every year since first officially reported by the Internet Crime Complaint Center (IC3) in 2000. Prior research studies and third-party reports of fraud show rates substantially higher than eBay’s reported negative feedback rate of less than 1%. The conclusion is most buyers are withholding reports of negative feedback.

Researchers Nikitov and Stone in a forensic case study of a single opportunistic eBay seller found buyers sometimes embedded negative comments in positive feedback as a means of avoiding retaliation from sellers and damage to their reputation. This category of positive feedback was described as “negative-positive” feedback. An example of negative-positive type feedback is “Good product, but slow shipping.”

This research study investigated the concept of using negative-positive type feedback as a signature to identify potential opportunistic sellers in an online auction population.

As experienced by prior researchers using data extracted from the eBay web site, the magnitude of data to be analyzed in the proposed study was massive. The nature of the analysis required - judgment of seller behavior and contextual analysis of buyer feedback comments – could not be automated. The traditional method of using multiple dedicated human raters would have taken months of labor with a correspondingly high labor cost. Instead, crowdsourcing in the form of Amazon Mechanical Turk was used to reduce the analysis time to a few days and at a fraction of the traditional labor cost.

The research’s results found that the presence of subtle buyer behavior in the form of negative-positive type feedback comments are an inter-buyer signal indicating that a seller was behaving fraudulently. Sellers with negative-positive type feedback were 1.82 times more likely to be fraudulent. A correlation exists between an increasing number of negative-positive type feedback comments and an increasing probability that a seller was acting fraudulently. For every one unit increase in the number of negative-positive type feedback comments a seller was 4% more likely to be fraudulent.

  Link to NovaCat

Share

COinS