Method and apparatus for multi-model hybrid comparison system
View Patent ↗A user authentication system collects measurements of physical and/or behavioral characteristics of a user. The measurements are processed by two or more processing engines to produce initial confidence measures, and a unified confidence measure is prepared from weighted inputs including the initial confidence measures.
1. A method comprising:
collecting a plurality of one type of biometric measurements of a user;
computing a first confidence measure from the biometric measurements according to a first algorithm and a second confidence measure from the biometric measurements according to a second algorithm,
wherein the first confidence measure is based on at least two of: a) string complexity based on a length of the string, repeated letters of the string, and repeated sequences of letters of the string, b) biometric template maturity based on a time elapsed since template creation, a time elapsed since the latest template update, a frequency with which the template has been used, and a frequency with which the template has been updated, and c) engine initial confidence based on an authentication scoring history of the user; and
wherein the second confidence measure is based on at least two of: a) string complexity based on a length of the string, repeated letters of the string, and repeated sequences of letters of the string, b) biometric template maturity based on a time elapsed since template creation, a time elapsed since the latest template update, a frequency with which the template has been used, and a frequency with which the template has been updated, and c) engine initial confidence based on an authentication scoring history of the user; and
preparing a unified confidence measure from weighed inputs including the first confidence measure and the second confidence measure.
2. The method of claim 1 wherein the plurality of biometric measurements includes a plurality of keystroke timings of the user typing a string.
3. The method of claim 2 wherein a weight associated with the first confidence measure incorporates at least one of a complexity measure of the string and the user's familiarity measure of the string.
4. The method of claim 2 wherein the string is one of a username, a real name, a password, an electronic mail address, or a predetermined phrase.
5. The method of claim 1 wherein the plurality of biometric measurements include at least one of a plurality of lengths of parts of the body of the user, an image of a fingerprint of the user, an image of an iris of the user, an image of a retina of the user, and a digitized recording of a voice of the user.
6. The method of claim 1 wherein the first confidence measure is an output of a neural network configured to classify an input data set, wherein classifying the input data set comprises learning that the user types overlaps successive keystrokes.
7. The method of claim 6 wherein the first confidence measure incorporates a maturity of a template.
8. The method of claim 6 , wherein classifying the input data set comprises learning that the user types repeated letters in a distinctive way.
9. The method of claim 1 wherein the second confidence measure is an output of a statistical analysis of an input data set.
10. The method of claim 1 further comprising:
updating a historical weight based on the first confidence measure, the historical weight to be used to prepare a subsequent confidence measure according to the first algorithm.
11. The method of claim 1 , wherein the plurality of one type of biometric measurements are only from a keyboard.
12. A non-transitory computer-readable medium containing instructions to cause a processor to perform operations comprising:
collecting one type of biometric measurements of a user;
processing the biometric measurements with a plurality of processing engines, each processing engine to produce an initial confidence result; and
computing a unified confidence result comprising a weighted sum of the initial confidence results, wherein weights of the weighted sum limit input of each processing engine to the weighted sum until confidence in performance of each processing engine is gained, and wherein confidence of each processing engine is based on an authentication scoring history of the user.
13. The non-transitory computer-readable medium of claim 12 , containing additional instructions to cause the processor to perform operations comprising:
normalizing the unified confidence result.
14. The non-transitory computer-readable medium of claim 12 , containing additional instructions to cause the processor to perform operations comprising:
invoking a plurality of plug-in modules found in a storage location.
15. The non-transitory computer-readable medium of claim 12 wherein a processing engine is to apply an algorithm to the biometric measurements to produce an initial confidence result.
16. The non-transitory computer-readable medium of claim 12 wherein collecting biometric measurements of a user comprises:
recording a time of a plurality of key press events and key release events, said key press and key release events associated with the user typing at least one of a username, a real name, a password, an electronic mail address, and a predetermined phrase.
17. A system comprising:
a processor;
a memory;
a storage device;
a hybrid pre-processor to collect one type of biometric input data;
a hybrid processor to pass a first dataset derived from the biometric input data to a first processing engine and to pass a second dataset derived from the biometric input data to a second processing engine; and
a hybrid unified score calculator to compute a unified confidence result from a first output of the first processing engine and a second output of the second processing engine,
wherein the first output of the first processing engine comprises a string complexity based on a length of the string, repeated letters of the string, and repeated sequences of letters of the string, and wherein the second output of the second processing engine comprises a string complexity based on a length of the string, repeated letters of the string, and repeated sequences of letters of the string.
18. The system of claim 17 wherein
the first processing engine is to apply a first algorithm to produce the first output based on the first dataset; and
the second processing engine is to apply a second algorithm to produce the second output based on the second dataset.
19. The system of claim 18 , wherein the first dataset is identical to the second dataset.
20. The system of claim 18 , wherein the first dataset is different from the second dataset.
21. The system of claim 17 wherein
the first processing engine is a statistical engine; and
the second processing engine is a neural network engine.