Method for generating a human likeness score

• US 8,776,173 B2
• Filed: 03/02/2012
• Issued: 07/08/2014
• Est. Priority Date: 03/24/2011
• Status: Active Grant

• Current Assignee: Imperva Incorporated (Thales SA)
• Original Assignee: AYAH Incorporated
• Inventors: Paxton, Tyler James, Tatoris, Reid Michael, Blackmer, Benjamin Crispell, Aleksa, Adam, VandenBrink, Stuart
• Primary Examiner(s): Patel, Nirav B
• Application Number:
  US13/411,071
• Publication Number:
  US 20120246737A1
• Time in Patent Office: 858 Days
• Field of Search: 726 26- 30, 726 2- 4, 726 16- 18, 382115-127
• US Class Current: 726/2
• CPC Class Codes: G06F 16/24578 using ranking
• CPC Class Codes: G06F 21/31 User authentication
• CPC Class Codes: G06F 21/316 by observing the pattern of computer usage, e.g. typical user behaviour
• CPC Class Codes: G06F 21/32 using biometric data, e.g. fingerprints, iris scans or voiceprints
• CPC Class Codes: G06F 21/36 by graphic or iconic representation
• CPC Class Codes: G06F 2221/2103 Challenge-response
• CPC Class Codes: G06F 2221/2133 Verifying human interaction, e.g., Captcha
• CPC Class Codes: H04L 63/083 using passwords cryptographic mechanisms or cryptographic arrangements for entity authentication using a predetermined code H04L9/3226

Abstract

One embodiment of the invention is a method utilizing a CAPTCHA to generate a human likeness score including blocks: a) receiving a user solution to the CAPTCHA; b) receiving a user interaction pattern descriptive of an interaction undertaken by the user, through a graphical interface of the CAPTCHA, to achieve the user solution; c) determining the accuracy of the user solution; d) comparing the user interaction pattern against an interaction model generated from interaction patterns of previous users; e) calculating the human likeness score based upon the determination of block c) and the comparison of block d), wherein the human likeness score lies within a continuum of human likeness scores.

Claims

1. A method for generating a human likeness score based on a Completely Automated Public Turing test to tell Computers and Humans Apart (CAPTCHA) provided through a graphical interface, the method comprising:

generating the CAPTCHA comprising the graphical interface and a first object and a second object depicted within the graphical interface;

pseudorandomly assigning a visual characteristic of the first object within the graphical interface to differentiate the CAPTCHA from a previously-generated CAPTCHA, the first object distinct from the second object by the visual characteristic;

assigning an instruction for solving the CAPTCHA based on the visual characteristic of the first object;

receiving a user solution to the CAPTCHA;

receiving a user interaction pattern descriptive of an interaction undertaken by a user, through the graphical interface, to enter the user solution;

determining an accuracy of the user solution in satisfying the instruction for solving the CAPTCHA;

comparing the user interaction pattern to an interaction model generated from interaction patterns of previous users; and

calculating the human likeness score based upon the accuracy of the user solution and the comparison of the user interaction pattern to the interaction model, the human likeness score lying within a continuum of human likeness scores.

2. The method of claim 1, further comprising comparing the human likeness score to a threshold likeness score to verify that the user is human.

3. The method of claim 2, further comprising adding the user interaction pattern to a database of user interaction patterns comprising interaction patterns of previous users.

4. The method of claim 2, wherein comparing the human likeness score to the threshold likeness score comprises determining the user to be human if the human likeness score is greater than the threshold likeness score.

5. The method of claim 2, wherein comparing the human likeness score to the threshold likeness score comprises determining the user to be automated software if the human likeness score is less than the threshold likeness score.

6. The method of claim 1, further comprising calculating an automated software likeness score based upon the accuracy of the user solution and the comparison of the user interaction pattern to the interaction model, the automated software likeness score lying within a continuum of automated software likeness scores.

7. The method of claim 1, wherein receiving the user interaction pattern comprises receiving an interaction selected from the group consisting of:

a cursor movement;

a mouse click;

a keystroke;

a voice command;

a joystick movement;

a touch;

a facial expression; and

a gesture.

8. The method of claim 1, further comprising receiving a set of user data distinct from the CAPTCHA, wherein calculating the human likeness score includes calculating the human likeness score further based upon the set of user data.

9. The method of claim 8, wherein the set of user data includes at least one of:

a CAPTCHA previously attempted by the user;

an Internet Protocol address of the user; and

a cookie associated with the user.

10. The method of claim 1, wherein calculating the human likeness score comprises implementing a decision tree to calculate the human likeness score.

11. The method of claim 1, wherein

determining the accuracy of the user solution comprises generating a first confidence group based upon the user solution;

comparing the user interaction pattern to the interaction model comprises generating a second confidence group based upon the user interaction pattern; and

calculating the human likeness score comprises determining the reliability of the first and the second confidence groups and compiling the first and the second confidence groups, based upon the determined reliability thereof, to generate the human likeness score.

12. The method of claim 11, further comprising:

accessing user data, including at least one of a CAPTCHA previously attempted by the user, an Internet Protocol address of the user, and a cookie associated with the user; and

generating a third confidence group based upon the user data;

wherein calculating the human likeness score further comprises determining the reliability of the third confidence group and compiling the first, the second, and the third confidence groups, based upon the determined reliability thereof, to generate the human likeness score.

13. The method of claim 1, wherein calculating the human likeness score comprises selecting a number value, from the continuum of human likeness scores comprising a continuum of number values, that correlates with a calculated confidence that the user is human.

14. The method of claim 1, further comprising generating the interaction model by aggregating interaction patterns of previous users, wherein the interaction model is indicative of a characteristic of an interaction undertaken by a human through the graphical interface.

15. The method of claim 1, wherein generating the interaction model comprises generating the interaction model from interaction patterns of a set of previous users comprising at least one known human.

16. The method of claim 1, further comprising isolating the user interaction pattern from the user interaction, wherein the user interaction comprises a path-type input provided through the graphical interface, the user interaction pattern comprises a direction change in a portion of the input path, and comparing the user interaction pattern to an interaction model comprises comparing the direction change of the user interaction pattern to the interaction model that defines a direction change typical of a human interaction.

17. The method of claim 1, further comprising isolating the user interaction pattern from the user interaction, wherein the user interaction comprises a path-type input provided through the graphical interface, the user interaction pattern comprises an input velocity of a portion of the input path, and comparing the user interaction pattern to an interaction model comprises comparing the input velocity of the user interaction pattern to the interaction model that defines an input velocity typical of a human interaction.

18. The method of claim 1, wherein pseudorandomly assigning the visual characteristic of the first object comprises pseudorandomly assigning at least one of an initial location, an initial orientation, an initial speed, and an initial direction of motion to the first object within the graphical interface.

19. The method of claim 1, wherein generating the CAPTCHA comprises generating the CAPTCHA in response to receiving a request, from a publisher, for the CAPTCHA.

20. The method of claim 19, wherein generating the CAPTCHA comprises receiving an asset from the publisher and incorporating the asset into the CAPTCHA as an advertisement.

21. The method of claim 1, wherein assigning the instruction for completing the CAPTCHA comprises assigning, to the CAPTCHA, an instruction related to at least one of the shape, the color, the size, the type, and the context of the first object within the graphical interface.

22. The method of claim 1, wherein receiving the user solution comprises receiving the user solution comprising manipulation of the position of the first object within the graphical interface.

23. The method of claim 22, wherein receiving the user solution comprises tracking manipulation of the first object within the graphical interface by the user.

24. The method of claim 23, wherein tracking manipulation of the first object within the graphical interface comprises recording a timing characteristic and a path characteristic of manipulation of the first object within the graphical interface by the user.

25. The method of claim 1, wherein generating the CAPTCHA comprises configuring the CAPTCHA for presentation within an interface selected from the group consisting of:

a website;

a mobile application;

a web-based forum; and

a web-based email client, and further comprising transmitting a portion of the CAPTCHA to a publisher for publication within the interface.

26. The method of claim 1, further comprising correlating the user solution with a user preference and adding the user solution to a catalog of user preferences corresponding to a survey associated with the CAPTCHA.

27. The method of claim 26, wherein adding the user solution to the catalog of user preferences comprises adding the user solution to the catalog in response to the human likeness score exceeding a threshold likeness score.

28. The method of claim 1, wherein receiving a user interaction pattern comprises receiving the user interaction pattern comprising a keystroke, and wherein comparing the user interaction pattern to the interaction model comprises comparing a state of a key and a timing of the keystroke to the interaction model.

29. A method for generating a human likeness score based on a Completely Automated Public Turing test to tell Computers and Humans Apart (CAPTCHA) provided through a graphical interface, the method comprising:

generating the CAPTCHA comprising the graphical interface and a first object and a second object depicted within the graphical interface;

assigning a visual characteristic of the first object within the graphical interface to differentiate the CAPTCHA from a previously-generated CAPTCHA, the first object distinct from the second object by the visual characteristic;

assigning an instruction for solving the CAPTCHA based on the visual characteristic of the first object;

receiving a user solution to the CAPTCHA, the user solution comprising a path-type input entered by a user to move at least one of the first object and the second object within the graphical interface to solve the CAPTCHA;

determining an accuracy of the user solution in satisfying the instruction for solving the CAPTCHA;

isolating a noise component from the path-type input of the user solution;

comparing the noise component to an interaction model, the interaction model defining input noise characteristics for human interactions and generated from path-type inputs entered by previous users to solve previous CAPTCHAs; and

calculating the human likeness score based upon the accuracy of the user solution and the comparison of the noise component to the interaction model, the human likeness score lying within a continuum of human likeness scores.

30. The method of claim 29, wherein isolating the noise component from the path-type input comprises calculating a signal-to-noise ratio of the path-type input, wherein calculating the human likeness score comprises calculating the human likeness score based on a deviation of the signal-to-noise ratio of the path-type input from a threshold signal-to-noise ratio defined by the interaction model.

31. The method of claim 30, wherein isolating the characteristic response to the object comprises extracting an input timing following a change in distance between the object and a second object moving within the graphical interface.

32. The method of claim 30, further comprising verifying that the user is human in response to the noise component exceeding the threshold signal-to-noise ratio and identifying the user as automated software in response to the noise component falling below the threshold signal-to-noise ratio.

33. A method for generating a human likeness score based on a Completely Automated Public Turing test to tell Computers and Humans Apart (CAPTCHA) provided through a graphical interface, the method comprising:

generating the CAPTCHA comprising the graphical interface and an object depicted within the graphical interface;

assigning a visual motion characteristic of the object within the graphical interface to differentiate the CAPTCHA from a previously-generated CAPTCHA;

assigning an instruction for solving the CAPTCHA;

receiving a user solution to the CAPTCHA, the user solution comprising a path-type input entered into the graphical interface by a user to solve the CAPTCHA;

determining an accuracy of the user solution in satisfying the instruction for solving the CAPTCHA;

isolating, from the path-type input, a characteristic response to the object moving within the graphical interface;

comparing the characteristic response to an interaction model, the interaction model defining human characteristic responses to objects moving within graphic interfaces and generated from path-type inputs entered by previous users to solve previous CAPTCHAs;

calculating the human likeness score based upon the accuracy of the user solution and the comparison of the characteristic response to the interaction model, the human likeness score lying within a continuum of human likeness scores.

34. The method of claim 33, further comprising verifying that the user is human in response to the human likeness score exceeding a threshold human likeness score and identifying the user as automated software in response to the human likeness score falling below the threshold human likeness score.

35. The method of claim 33, wherein isolating the characteristic response to the object comprises isolating a lag in direction change of the path-type input relative to a direction change of the object moving within the graphical interface, wherein calculating the human likeness score comprises calculating the human likeness score based on a comparison of the lag in direction change of the path-type input to the interaction model defining lags in direction changes of path-type inputs by humans.

36. The method of claim 33, wherein isolating the characteristic response to the object comprises identifying an input path of a cursor leading up to selection of the object moving within the graphical interface, wherein calculating the human likeness score comprises calculating the human likeness score based on a comparison of the input path of the cursor to the interaction model defining input paths of cursors leading up to selections of objects within graphical interfaces by humans.

37. The method of claim 33, wherein isolating the characteristic response to the object comprises identifying an input path of a cursor following release of the moving object within the graphical interface, wherein calculating the human likeness score comprises calculating the human likeness score based on a comparison of the input path of the cursor to the interaction model defining input paths of cursors following releases of objects within graphical interfaces by humans.