'Viceroi' algorithm improves detection of click fraud
A group of researchers have devised an algorithm they say could help advertising networks better detect fraudulent clicks.
Fraudsters have developed sophisticated ways to perpetrate click fraud, which involves using various methods to generate fake clicks on advertisements, defrauding advertisers. Digital marketing revenues are rapidly growing and exceeded US$36 billion in 2012 in the U.S., according to the Interactive Advertising Bureau.
Advertising networks are secretive about the technologies they use to stop click spam. Often, it involves filtering out attacks, such as if thousands of clicks on an advertisement are coming from a single IP address. But defensive moves still miss attacks, wasting advertisers' money.
The researchers' algorithm, called Viceroi, is free and can be used by advertising networks. Viceroi looks for publishers who have abnormally high per-user revenues, which may be an indication of fraud. For their experiment, Viceroi was tested with a major ad network, flagging several hundred publishers as suspects out of tens of thousands, according to their research paper.
Vacha Dave, a post-doctoral researcher at the University of California at San Diego and co-author of the paper, said in interview Thursday that per-user revenue rates at some publishers were way higher than those collected by Google or Microsoft.
Viceroi works because of the economics of click spam. In one variation of the fraud, a click spammer may pay someone else a per-install fee to distribute a dodgy search toolbar designed to direct people to their advertisements.
The toolbar's search results page is stuffed with advertisements since the click spammer wants to exploit the user as much as possible before the tool is uninstalled. But the rising per-user revenue on a publisher's site would be spotted by Viceroi.
To beat Viceroi, the "click spammers must reduce their per-user revenue to that of an ethical publisher. At which point, without the economic incentive to offset the risk of getting caught, the net effect is a disincentive to commit click spam," the paper said.
Not all publishers are necessarily at fault if they have abnormally high per-user revenues. There is a lot of traffic brokering on the Internet, and it is often hard to tell where user traffic originated from, said Saikat Guha of Microsoft Research India, who co-authored the paper. Advertising networks learn from Viceroi which publishers to investigate.
"Some of the publishers are definitely being take advantage of," Guha said. "Our job is to help them find the bad traffic."
The research paper, also authored by Yin Zhang of the University of Texas at Austin, will be presented at the ACM Conference on Computer and Communications Security in Berlin, which will be held Nov. 4-8.