AJ Guardado Comp Sci 49S Eddie Morgan Scribe Notes In 2002, Google added AdWords, as well as three different filters to detect the invalid clicks made on this program, which they would keep track of over time. Click fraud was still new to Google at this time, and Google was still having problems from changing pay models. The next year, Google debuted AdSense and formed the Click Quality Team, as well as offline protection methods. Even with all these resources, Google could not stop the more complicated and sophisticated attacks from users. By 2005, Google believed they had most of their problems with fraudulent clicks worked out. Added new software to combat fraud, and also revised the detection and determination of double clicks because advertisers were angry at having to pay double for innocent clicks by the same person. Some accused Google of having “biased” filters, but Google reasoned they gained no advantage by doing this. Google feared they would drive their advertisers away by doing this, and would lost more by not protecting the money of their advertisers. Click fraud is now handled in the engineering department of Google, completely separate from the finance department. Offline detection is a more extensive detection process, because there are no time restraints in which to analyze clicks. 1 AJ Guardado Comp Sci 49S Detection can be made by alerts or auto-termination, and are used for detecting more subtle or advanced clicks that are then looked over by the Click Quality Team, who try to determine if the click is valid or invalid. Auto-termination warns, investigates or terminates a published based on the severity of that particular incident. Minor incidents will be warned, but more severe cases, or multiple minor incidents will result in auto-termination by Google. Detection can now manually be done by the Click Quality Team or an automated program run by Google. Google is not making good enough use of data mining methods which help to determine if clicks are valid or invalid with greater accuracy. The biggest problem with this is that they need to build a program that classifies clicks properly, and then train all of the Team employees to use this program. Also not using the conversion data collected to its full extent; mostly because a small amount of the clicks made have information which occurs after a specific time the click has been made. Google should look into the future, purchase more advanced filters, and get ready for the next level of fraudulent users that is sure to come in the future. 2