Click-Fraud Resistant Methods for Learning Click-Through Rates
by Nicole Immorlica
In pay-per-click online advertising systems like Google, Overture, or MSN, advertisers are sold online advertisement slots in an auction and then charged for their ads only when a user clicks on the ad. While these systems have many advantages over other methods of selling online ads, they suffer from one major drawback. They are highly susceptible to a particular style of fraudulent attack called click fraud. Click fraud happens when an advertiser or service provider generates clicks on an ad with the sole intent of increasing the payment of the advertiser. Leaders in the pay-per-click marketplace have identified click fraud as the most significant threat to their business model. In this talk, we see that in auction mechanisms based on estimating the click-through rates (or probability that an advertisement receives a click), a fraudulent click causes not only a short-term loss, but also a long-term gain due to an increase in the estimate of the click-through rate. We prove that for a particular class of click-through rate learning algorithms, these two effects cancel, thereby reducing click fraud to impression fraud. This work leads to an exciting new research direction, that of learning in the presence of noise manipulated by strategic agents.
This is joint work with Jain, Mahdian, and Talwar.
Back to the LCI Forum page