With the state of 2008 the global net advertising revenues are at the level of 31.7 billion € with an increasing amount of 31%. The current estimations for 2016 are about 62 billion € - just for the US ad market. Google Inc, the leading company in this topic estimates their fraud rate at about 30% of the total traffic.
So, what we notice is a huge market with much potential and an increasing dark side of click fraud.
What is click fraud?
Click fraud occurs on the Internet in pay per click online advertising when a person, automated script or computer program imitates a legitimate user of a web browser clicking on an ad, for the purpose of generating a charge per click without having actual interest in the target of the ad's link. Therefore the advertisers have to pay unjustifiable and publishers earn due to a service that was not made. Furthermore the advertisers do not just make a financial loss by the fee but also by the aspect that their online marketing campaign ends in smoke. Hence, fraud is not only about profits but also about vandalism.
But what’s all about this?
I am currently developing a fraud detection and prevention concept for the Plista GmbH in order to deal with the issue of click fraud. There are many aspects that have to be observed, but for a given occasion, I would like to focus on the more fundamental motivations of fraudsters.
During the conceptual development I subdivided the involved publishers in three types. As one is able to see, the priority and cause of the involved publisher varies from type to type.
|
Type |
Cause |
Priority for Publishers |
|
passive unknowingly |
through crawler or malevolent competitors |
his website |
|
passive knowingly |
Ego of the webmaster, statspedding, triviality |
his website |
|
active |
Fraud as a major business, base motives |
his earnings through fraud |
behavioral analysis against fraud
Each of these types reflects a different characteristic and its own pattern of behavior, which were set up based on the Broken Windows Theory. This theory is the foundation of the so-called zero tolerance strategy. It is a principle rather than a crime control theory, since it does not explain the causes of crime, but merely describes symptoms. In addition it states that maintaining and monitoring urban environments in a well-ordered condition may stop further vandalism and escalation into more serious crime. Nevertheless, it serves as a research base today many favorite crime-prevention measures.
|
Type |
Characteristic |
Behavior Pattern |
|
passive unknowingly |
broad-minded, influenceable, reasonable, need for agreement |
cooperates, because of insight and a poor basis for negotiation |
|
passive knowingly |
“bumble about”, influenceable unpredictable, ego usually less than long-term profitability |
agrees to a deal (not counting Fraud + admonishment in return for disbursement), intimidation by Legal possible |
|
active |
technically gifted and calculative, purposeful, knows what he's doing, predictable, unswayable, often moves in the gray / black zone |
potentially no contact since risk and additional expense (new account < Communication) |
Due to this behavioral analysis we are now able to establish detection and prevention measures against each of the individual types in order to prevent them. Like we are able to see, the consultation in case of unknowingly fraud would be beneficial for both sides. In case of knowingly fraud it won’t, because the fraudster would never want to admit his involvement and thus to forego part of his profit. So we have to figure out different methods in order to handle knowingly fraud. Since these measures effectively influence the traffic quality of the Plista platform i am only able to reveal the usage of the predictable behavioral targeting concept as well as the collaborative filtering principal at this point in order to solve this initial point. Some example methods are described in Dr. Alexander Tuzhilin Report about Google AdSense.
Read more about our Anti Fraud Pattern by Framsteg Think Tank
