With the popularity of online advertising, the mode of Pay by Per Click is gradually accepted by everyone. However, the problem that follows is that Fraud Clicking is under the eyebrows, as this will directly be related to whether such ad mode can survive and can become a real website owner's income source.
Here is how the Google Adsense system prevents clicking spoof from the perspective of the system, and hopes to prevent false click to prevent false click to prevent false clicks:
1] Click rate = number of clicks / total views. The click rate is a key way to determine if there is a keyless way, you can imagine that the clicks on a website exceeds 10% of this.
# o f / / / /
2] Click over the coverage / independent IP, this distribution is if there is; a single IP (click / Browse) = Click the coverage rate of more than three times the system error range will have cheating.
For example, for example from 129.119.200.1, users have viewed 16 web pages, clicked 4 ads, and the click rate of the entire ad "" to "to" in [1] is 5%, then calculate:% 5 x 16 = ~ 1, the variance is SQRT (1) = 1, click the coverage rate = 4/1 = 4, according to the math, this probability is less than one thousandth.
Ratio VS IP Distribution
3] Clicking Rate "Click Cable Rate" / IP / Time to analyze the click rate according to the time series, if there is a clear peak on a certain period of time, then this will think that there is a potential spoofing clicks.
Ratio vs Time
4] Time and advertisement of web pages LOAD Click time difference analysis, and analysis of time difference sequences between clicks [Page LOAD time and advertising Click time difference] should be a Poisson distribution Possion Distribution, and between each cliff The time difference should also be a Possion Distribution. If this time is used in seconds, the shape of the Gaussian distribution is substantially presented by more than 25 seconds.
[Time of Loading - Time of Click] Distribution vs Possion [Time Difference of Two Click] Distribution vs Possion / Gaussion
5] In response to Proxy Click to change IP Click to say that in the past, it is the most difficult to solve the most difficult to find cheating method. When the Chinese people conduct Alexa's Boost, I use proxy to fake false click, but here as long as they are reverse Check if IP's source is a server with Proxy functionality.
Reverse proxy check
6] Analysis of HTTP_Agent analysis HTTP_AGENT / time analysis, peak exceeding 3 variance needs to be review
7] Analysis of HTTP_REFERRAL analysis of the time series of the Referral / Time, the peak value is more than 3 variance needs to be reviewed
8] The overall effect is also a very useful amount: all users' validity of each user / independent IP will be able to find the spam clicking running computer more directly and block.
Overall Ratio VS IP
Even if I have given the above ways to prevent cheating, don't forget:
Evil people are always more than just people.
