Content area
Full Text
Abstract: As a result of consistent double-digit year-over-year growth rates in e-commerce sales, marketing firms continue to aggressively seek better means to aid in classifying user's cyber behaviors, thereby improving personalization, product recommendation and prediction. While motivated purely by financial incentives, this type of work has provided great insights into the type of information which may be gained from user's online activities, and has made significant strides in cyber-based behavioral modeling. An aspect of this type of research receiving much less attention focuses on whether these cyber behaviors are descriptive enough to uniquely identify an individual user. While the ability to uniquely identify individuals based on their online activities has e-commerce ramifications, its greatest potential may be in the security realm. The ability to create a "cyber fingerprint" of an individual to uniquely distinguish them provides a baseline from which variations in behavior may be identified. Such a mechanism could then be used to detect and prevent insider threat, fraud, and hacker activity by triggering alerts when a user behaves in a manner inconsistent with their previously established "norm". In this paper, we investigate whether a user's search activities provide an accurate model for the identification of a user and propose a formal approach to calculate the minimal amount of data required to create such a model. We make use of three months worth of real world query logs and apply a supervised learning algorithm to ascertain whether users can be discriminated through search queries alone. Experimental results are provided demonstrating the effectiveness of our fingerprinting technique and sample size estimation methods. Finally, the implications of this research in areas such as e-commerce, insider threat detection, and fraud detection are discussed.
Keywords: User modeling, cyber fingerprint, behavior, meta-activity
(ProQuest: ... denotes formula omitted.)
1. Introduction
While individuals like to see themselves as free-willed, self-governing, and unique, research suggests the exact opposite may be true. Work done by Eagle et al (Eagle 2006) tracking the geographic location of college students found approximately ninety percent of what most people do in any day follows routines so complete, that their behavior can be predicted with alarming accuracy. Similar research (Catledge 1995, Cockburn 2001, Herder 2005) has shown that a "creature of habit" mentality also takes place...