Massive levels of click fraud

A report from Anchor Intelligence puts current levels at 35% in the US and 35.2% in Australia… that seems like massive levels to me…

I haven’t read the entire PDF report but the write-up in Webpronews seems to suggest that it is all highly organised…

How do they do it? I would have thought unusual levels of click throughs would produce patterns that are easy to spot

http://www.webpronews.com/topnews/2010/04/08/attempted-click-fraud-rate-surges-in-q1

http://www.anchorintelligence.com/trafficqualityreport_q12010.pdf

If the bad activity is so common (35% is more than 1 in 3 clicks), and hard to identify even for a human reviewing the activity (botnets are your computer and mine, taken over silently, so their activity really is a browser clicking a link and won’t leave any signs in the HTTP requests)… then you lose the ability to train machine learning algorithms to find such patterns. There is no clean sample to teach the algorithm the difference between valid and malicious patterns.

It sounds like a “hard problem” in the computer science sense.

this has been a problem for many years now, i remember i had this conversation (botnet click fraud) about 5yrs ago with a friend.

and this is just one of the things a botnet can be used for. they also switch botnets with each others, and in other cases also rent them out.

as Dan said above here, this botnets are computers like your and mine that has been taken over silently, and turned into a zombie. which they can use for just about anything. it is a lot of different computers, in different countries, with different operating systems and different ip’s… so they are running stealth, and really not easy to spot.

many of these botnets are controlled from IRC, where they get their commands and missions, from the owners.