Detecting frauds in online advertising systems

Abstract

Online advertising is aimed to promote and sell products and services of various companies in the global market through internet. In 2005, it was estimated that companies spent $10B in web advertisements, and it is expected to grow by 25-30% in the next few years. The advertisements can be displayed in the search results as sponsored links, on the web sites, etc. Further, these advertisements are personalized based on demographic targeting or on information gained directly from the user. In a standard setting, an advertiser provides the publisher with its advertisements and they agree on some commission for each customer action. This agreement is done in the presence of Internet Advertising commissioners, who represent the middle person between Internet Publishers and Internet Advertisers. The publisher, motivated by the commission paid by the advertisers, displays the advertisers’ links in its search results. Since each player in this scenario can earn huge revenue through this procedure, there is incentive to falsely manipulate the procedure by extracting forbidden information of the customer action. By passing this forbidden information to the other party, one can generate extra revenue. This paper discusses an algorithm for detecting such frauds in web advertising networks.