Click fraud detection

• US 8,103,543 B1
• Filed: 01/27/2010
• Issued: 01/24/2012
• Est. Priority Date: 09/19/2006
• Status: Active Grant

• Current Assignee: Gula Consulting Limited Liability Company (Intellectual Ventures LLC)
• Original Assignee: GERE DEV Applications LLC (Intellectual Ventures LLC)
• Inventors: Zwicky, Richard Kazimierz
• Primary Examiner(s): ROBINSON, GRETA LEE
• Application Number:
  US12/694,706
• Time in Patent Office: 727 Days
• Field of Search: 707/687, 707/688, 707/690, 707/705, 707/999.107, 709/224, 705/50, 705/52, 705/64, 705/65, 705/72, 705/74, 705/75, 705/14.15, 705/14.26, 705/317, 705/318
• US Class Current: 705/14.26
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0225 Avoiding frauds
• CPC Class Codes: G06Q 30/0246 Traffic
• CPC Class Codes: G06Q 30/0248 Avoiding fraud

Abstract

Systems and methods for detecting instances of click fraud are disclosed. Click fraud occurs when, for example, a user, malware, bot, or the like, clicks on a pay per click advertisement (e.g., hyperlink), a paid search listing, or the like without a good faith interest in the underlying subject of the hyperlink. Such fraudulent clicks can be expensive for an advertising sponsor. Statistical information, such as ratios of unpaid clicks to pay per clicks, are extracted from an event database. The statistical information of global data is used as a reference data set to compare to similar statistical information for a local data set under analysis. In one embodiment, when the statistical data sets match relatively well, no click fraud is determined to have occurred, and when the statistical data sets do not match relatively well, click fraud is determined to have occurred.

Claims

1. A computer-implemented method for detecting click fraud, the method comprising:

extracting activity data from an event database;

generating statistical information of unpaid referrals relative to paid referrals for a reference data set;

generating statistical information for a data set under analysis;

comparing the statistical information for the reference data set and the statistical information for the data set under analysis; and

assessing whether or not click fraud is present based at least partly on the comparison;

wherein said extracting, generating, comparing, and assessing are performed by one or more computing devices;

wherein generating statistical information of unpaid referrals relative to paid referrals for the reference data set comprises computing ratios of unpaid referrals to paid referrals and computing at least one of standard deviation, skewness, or kurtosis of the computed ratios for the reference data set;

wherein generating statistical information for the data set under analysis comprises computing ratios of unpaid referrals to paid referrals and computing at least the one of standard deviation, skewness, or kurtosis of the computed ratios for the data set under analysis;

wherein the data set under analysis comprises data collected over a shorter period of time than data for the reference data set.

2. The method of claim 1, wherein the reference data set corresponds to about a week of data, and wherein the data set under analysis corresponds to about an hour of data.

3. The method of claim 1, wherein the paid referrals correspond to pay per click referrals.

4. The method of claim 1, wherein the paid referrals correspond to pay per impression referrals.

5. An apparatus for detecting click fraud, the apparatus comprising:

a descriptive statistics system comprising computer hardware, the descriptive statistics system configured to;

extract activity data from an event database;

generate statistical information of unpaid referrals relative to paid referrals for a reference data set; and

generate statistical information for a data set under analysis; and

an inferential statistics system configured to;

compare the statistical information of the reference data set and the statistical information of the data set under analysis; and

assess whether or not click fraud is present based at least partly on the comparison;

wherein the descriptive statistics system is further configured to compute ratios of unpaid referrals to paid referrals and compute at least one of standard deviation, skewness, or kurtosis of the computed ratios for the reference data set to generate statistical information of unpaid referrals relative to paid referrals for the reference data set;

wherein the descriptive statistics system is further configured to compute ratios of unpaid referrals to paid referrals and compute at least the one of standard deviation, skewness, or kurtosis of the computed ratios for the data set under analysis to generate statistical information for the data set under analysis;

wherein the data set under analysis comprises data collected over a shorter period of time than data for the reference data set.

6. The apparatus of claim 5, wherein the reference data set corresponds to data for about a week, and wherein the data set under analysis corresponds to data for about an hour.

7. The apparatus of claim 5, wherein the paid referrals correspond to pay per click referrals.

8. The apparatus of claim 5, wherein the paid referrals correspond to pay per impression referrals.

9. A computer readable medium having instructions stored thereon for detecting click fraud, the instructions comprising:

instructions to extract activity data from an event database;

instructions to generate statistical information of unpaid referrals relative to paid referrals for a reference data set;

instructions to generate statistical information for a data set under analysis;

instructions to compare the statistical information for the reference data set and the statistical information for the data set under analysis; and

instructions to assess whether or not click fraud is present based at least partly on the comparison;

wherein the instructions to generate statistical information of unpaid referrals relative to paid referrals for the reference data set comprise instructions to compute ratios of unpaid referrals to paid referrals and instructions to compute at least one of standard deviation, skewness, or kurtosis of the computed ratios for the reference data set;

wherein the instructions to generate statistical information for the data set under analysis comprise instructions to compute ratios of unpaid referrals to paid referrals and instructions to compute at least the one of standard deviation, skewness, or kurtosis of the computed ratios for the data set under analysis;

wherein the data set under analysis comprises data collected over a shorter period of time than data for the reference data set.

10. The computer readable medium of claim 9, wherein the reference data set corresponds to about a week of data, and wherein the data set under analysis corresponds to about an hour of data.

11. The computer readable medium of claim 9, wherein the paid referrals correspond to pay per click referrals.

12. The computer readable medium of claim 9, wherein the paid referrals correspond to pay per impression referrals.