AdSense fraud system - a guess

Pure speculation

[darkmage]

Hi All

I have seen lots of posts about Google's fraud checking program. Here are some ideas as to how it could work.

Firstly, Google would need to record each ad impression, URL, IP, time etc for auditing the adsense program.

Google would record similar data when you log in to your adsense account

Most log files record all sorts of info about a computer - version of Windows, browser etc. These help build a profile/fingerprint of the user.

I am sure Google would have referral stats. If you go to your own site via bookmark or typed URL, and click an ad, I reckon this is a red flag. Match this to your login details (assuming same IP address) and the score builds.

Multiple clicks from the same site on the same IP address is a big red flag to the whole adsense program, so I am sure the system is sniffing them out.

Importantly, there is the wonderfully simple idea of your own cookies used to login. This could give you away in a flash.

This is not a definitive list, but Google seems to like build a complex matrix for everything it does, so each item can add or subtract to the fraud score. I suspect that it also multiplies up. One click is probably ignored, 2 is weighted four times as bad etc.

In short, you'd have to be pretty dumb to even try clicking, even once. I would be a little cautious about using public computers to log in to your account, but really what are the odds that someone uses that computer, 'accidentially' visits your site and clicks an ad mutiple times? If it concerns you, flush the cookies, cache and history. Shared home or work computers may require some education of the other users.

DM

[danny]

The most important information Google has for detecting fraud is the quite general information it has about web use. Between the toolbar and search stats, it has enormous amounts of quite general information on what web sites are visited by how many people how often and when, and this provides the background against which patterns in ad clickthroughs can be evaluated.

I wouldn't be at all surprised if the IP addresses of AdSense publisher logins aren't used at all, or are only checked if something else flags the attention of a human.

[dhatz]

[webmasterworld.com...]

please read msg #19

[ChrisKud5]

If you think first about the number of publishers, and then about the number of pages with adsense on them, and then the daily traffic to those pages, or impressions, you are looking at tens of hundreds of millions of impressions per day for ads (or whatever large number it actually is!)

A record with url, ip, os, broswer, etc, etc etc is a HUGE HUGE amount of data. How long can data storage for this go on?

I find it very hard to belive that data storage and recording goes on for each and every impression and each and every log in done by adsense impressions and client log ins. The fact is it is just a huge amount of data and not one that can really be stored and searched through in real time throughout a day week month year etc.

Sure it is possible but the sheer size of the data and processing power required to update the data and search through ALL RECORDS for clicks that happened from the same IP on an IP that someone is logging in on.

The whole science behind the thing is why google employees get paid and we don't. The amount of data and searching through data that goes on is out of this world.

[Bluepixel]

They are certainly logging it since they are logging everything since they started there search service.
If google discovers a clickbot in the future, it can analyze it and if it finds any design flaws, rescan the logs and ban advertisers which cheated in the past.

I would do that if I were google :-).