Method and system for validating personalized account identifiers using biometric authentication and self-learning algorithms

• US 10,650,379 B2
• Filed: 06/24/2013
• Issued: 05/12/2020
• Est. Priority Date: 03/26/2013
• Status: Active Grant

• Current Assignee: TATA Consultancy Services Limited (Tata Sons Pvt Ltd.)
• Original Assignee: TATA Consultancy Services Limited (Tata Sons Pvt Ltd.)
• Inventors: Agrawal, Sachin Kumar, Garg, Shalin, Vallat, Sathish
• Primary Examiner(s): McAtee, Patrick
• Assistant Examiner(s): Martinez-Hernandez, Edgar R
• Application Number:
  US13/925,307
• Publication Number:
  US 20140297528A1
• Time in Patent Office: 2,514 Days
• CPC Class Codes: G06F 21/32 using biometric data, e.g. fingerprints, iris scans or voiceprints
• CPC Class Codes: G06Q 20/40145 Biometric identity checks
• CPC Class Codes: G10L 17/24 the user being prompted to utter a password or a predefined phrase

Abstract

Disclosed is a system and method for biometric authentication of a user using a personalised identification and associated biometric data. In one embodiment, a plurality of personalised identifiers and biometric data may be captured from a number of users and stored in a repository as stored records. The authentication process may be divided into two phases. In the first phase, either a speech recognition or character recognition process may be applied in order to determine the text spoken or written by the user. Subsequently a few records may be fetched from the repository on the basis of text mapping. In the second phase, biometric authentication may be performed by comparing the biometric sample with the stored biometric data corresponding to the fetched records to uniquely identify a single user. Further a machine learning technique may be applied in order to periodically refine a plurality of models stored in the repository.

Claims

1. A computer implemented method for identifying and authenticating at least one user by means of at least one personalized identifier, the method comprising:

storing, by a processor, a first set of records associated with the at least one user, each record comprising a text, a golden copy of biometric templates and a model associated with the golden copy of biometric templates, wherein the text, the golden copy of biometric templates and the model are associated with the at least one personalized identifier;

assigning, by the processor, at least one privilege level to the at least one personalized identifier in each record of the first set of records based on one or more attributes;

capturing, by the processor, the at least one personalized identifier in the form of a first real time biometric sample wherein the first real time biometric sample is in the form of speech or a handwriting signature;

converting, by the processor, the first real time biometric sample into one or more text-scripts;

retrieving, by the processor, a second set of records from the first set of records, by matching the one or more text-scripts with the stored text in each record of the first set of records and accordingly identify a model set comprising one or more models associated with each record from the first set of records and the second set of records;

comparing, by the processor, a biometric template corresponding to the biometric sample with the golden copy of biometric templates of each of the records from the second set of records to generate a matching score, wherein the matching score is generated for each model in the model set;

identifying, by the processor the at least one user uniquely from the model set having the matching score greater than or equal to a pre-defined threshold, when the number of models in the model set having the matching score greater than or equal to a pre-defined threshold score is more than one;

prompting, by the processor, the at least one user to provide a second real time biometric sample corresponding to a randomly generated text displayed to the at least one user, wherein the second real time biometric sample is in the form of a speech or a handwriting signature;

comparing, by the processor, a biometric template corresponding to the second real time biometric sample with the golden copy of biometric templates associated with the models from the model set, having the matching score greater than or equal to a pre-defined threshold score to uniquely identify the at least one user;

re-calibrating, by the processor, the model set having the matching score greater than or equal to a pre-defined threshold score using the biometric template corresponding to the second real-time captured biometric sample, wherein the model set is being created and recalibrated using at least one machine learning technique selected from a group comprising at least one of a decision tree learning, an association tree learning, and an Artificial neural network based on the second set of records; and

generating, by the processor, a user dependent model using the at least one machine learning technique and metadata of the at least one user, wherein the metadata is used for identification of the at least one user from a set of users.

2. The method of claim 1, further comprising:

determining that a number of models in the model set having matching score greater than or equal to the pre-defined threshold score is more than one;

prompting, by the processor, the user to provide a real-time biometric sample corresponding to a randomly generated text displayed for the user;

comparing, by the processor, a biometric template corresponding to the real-time biometric sample with the golden copy of biometric templates associated with the models from the model set having matching score greater than or equal to the pre-defined threshold score; and

re-calibrating, by the processor, the models having matching scores greater than or equal to the pre-defined threshold score using the biometric template corresponding to the real-time captured biometric sample.

3. The method of claim 1, wherein the at least one personalized identifier is a variable length data selected from a group comprising a numeric data, an alphabetic data and combinations thereof.

4. The method of claim 3, wherein the at least one personalized identifier is capable of being either written or spoken, and capable of being transformed from either image to text or speech to text, respectively.

5. The method of claim 1, wherein the captured personalized identifier is converted into the one or more text-scripts using speech-to-text conversion or Optical Character Recognition (OCR) techniques.

6. The method of claim 1, wherein the text is converted into a hash-code or Unicode data format of standard data size.

7. The method of claim 1, wherein the at least one personalized identifier is captured in the form of a non-transcriptable sounds, and is correlated with stored text after removing and suppressing the non-transcriptable sounds.

8. The method of claim 1, wherein the matching of the one or more text-scripts with the text is performed using at least one algorithm selected from a group comprising:

a semantic string matching, an approximate string matching, a fuzzy string matching, and combinations thereof.

9. The method of claim 2, wherein the model set is being created and recalibrated using at least one machine learning algorithm selected from a group comprising:

a Decision tree learning, an Association rule learning, an Artificial neural network, Genetic programming, Inductive logic programming, Support vector machines, Clustering, Bayesian networks, Reinforcement learning, Representation learning, Similarity and metric learning, Sparse Dictionary Learning, and combinations thereof.

10. The method of claim 2, wherein the model is created and re-calibrated utilizing metadata extracted from the real-time biometric sample and the golden copy of biometric templates.

11. The method of claim 1, wherein the one or more attributes are selected from a group comprising transaction location, transaction purpose, transaction limit, network type, duration of transaction, and combinations thereof.

12. A computer implemented system for identifying and authenticating at least one user by means of at least one personalized identifier, the system, comprising:

a processor; and

a memory coupled to the processor, wherein the processor is configured to;

store, a first set of records associated with the at least one user, each record comprising a text, a golden copy of biometric templates and a model associated with the golden copy of biometric templates, wherein the text, the golden copy of biometric templates and the model are associated with the at least one personalized identifier;

assign, at least one privilege level to the at least one personalized identifier in the each record of the first set of records based on one or more attributes;

capture the at least one personalized identifier in the form of a first real time biometric sample, wherein the first real time biometric sample is in the form of speech or a handwriting signature;

convert the first real time biometric sample into one or more text-scripts;

retrieve a second set of records from the first set of records, by matching the one or more text-scripts with the text in each record of the first set of records and accordingly identify a model set comprising one or more models associated with each record from the first set of records and the second set of records;

compare a biometric template corresponding to the biometric sample with the golden copy of biometric templates of each of the records from the second set of records to generate a matching score, wherein the matching score is generated for each model in the model set;

identify the at least one user uniquely using the model from the model set having the matching score greater than or equal to a pre-defined threshold, when the number of models in the model set having the matching score greater than or equal to a pre-defined threshold score is more than one;

prompt, the at least one user to provide a second real time biometric sample corresponding to a randomly generated text displayed to the at least one user, wherein the second real time biometric sample is in the form of a speech or a handwriting signature;

compare a biometric template corresponding to the second real-time biometric sample with the golden copy of biometric templates associated with the models from the model set, having the matching score greater than or equal to a pre-defined threshold score to uniquely identify the at least one user;

re-calibrate the model set having the matching score greater than or equal to a predefined threshold score using the biometric template corresponding to the second real-time captured biometric sample, wherein the model set is being created and re-calibrated using at least one machine learning technique selected from a group comprising at least one of a decision tree learning, an association tree learning, and an Artificial neural network based on the second set of records; and

generate a user dependent model using the at least machine learning technique and metadata of the at least one user, wherein the metadata is used for identification of the at least one user from a set of users.

13. The system of claim 12, the instructions further comprising instructions for:

determining that a number of models in the model set having matching score greater than or equal to the pre-defined threshold score is more than one;

prompting the user to provide a real-time biometric sample corresponding to a randomly generated text displayed for the user;

comparing a biometric template corresponding to the real-time biometric sample with the golden copy of biometric templates associated with the models from the model set having matching score greater than or equal to the pre-defined threshold score; and

re-calibrating the models having matching score greater than or equal to the predefined threshold score using the biometric template corresponding to the real-time captured biometric sample.

14. The system of claim 12, wherein the text-scripts are converted to text wherein the text-scripts to text conversion comprises:

speech-to-text conversion or optical character recognition (OCR) for converting speech-to-text, image-to-text and handwriting-to-text respectively.

15. The system of claim 13, the instructions further comprising instructions for:

assigning, by the processor, at least one privilege level to the at least one personalized identifier in each record.