Systems and methods of photo-based fraud protection

Patent metadata

Field	Value
Publication number	US-10861025-B2
Application number	US-201815910330-A
Country	US
Kind code	B2
Filing date	Mar 2, 2018
Priority date	Mar 2, 2018
Publication date	Dec 8, 2020
Grant date	Dec 8, 2020

Abstract

A computer-implemented method of fraud protection is described. A server may receive one or more keywords and a likelihood of fraud from an external source. The server may receive an image of a document from a user device, wherein the document comprises at least one of a written communication or a printed communication. The server may extract text from the image of the document, compare the extracted text to the one or more stored keywords, and calculate a confidence level of fraud. The server may send an indication of the confidence level of fraud to the user device.

First claim

What is claimed is:

1. A computer-implemented method of fraud protection, comprising:
scraping, by a server, an image of a physical piece of mail in association with actual mail delivery from a user's electronic messaging account;
extracting, by the server, at least one of graphics and text from the image, the text being extracted using optical character recognition;
comparing, by the server, the extracted text to a database of keywords, wherein each keyword has an associated likelihood of fraud determined via a machine learning algorithm trained on instances of known fraud obtained from at least one database maintained by at least one of a financial institution, a government agency, or a law enforcement agency;
determining, by the server, at least one word from the extracted text that matches a keyword in the database;
calculating, by the machine learning algorithm, a confidence level of fraud for the physical piece of mail based on the likelihood of fraud of the at least one determined word that matches and a frequency of appearance in the image of the at least one determined word that matches; and
in response to calculating a confidence level of fraud higher than a pre-defined threshold, communicating, by the server, an indication to display a notification of the confidence level of fraud through a user interface of a user device associated with the user.

2. The method of claim 1 , comprising:
receiving one or more reports of potential fraud from one or more users; and
updating the one or more keywords and the likelihood of fraud associated with the one or more keywords based on the received one or more reports.

3. The method of claim 1 , wherein the extracted graphics comprise at least one of a logo, a trademark, or a signature.

4. The method of claim 1 , comprising:
obtaining a name of a sender from the extracted text;
receiving a first group of contact information associated with the name of the sender from an external source, wherein the first group of contact information comprises at least one of a postal address, a physical address, a phone number, or an email address;
obtaining a second group of contact information of the sender from the extracted text, wherein the second group of contact information comprises at least one of a postal address, a physical address, a phone number, or an email address;
comparing the second group of contact information to the first group of contact information; and
updating the confidence level of fraud based on comparing the second group of contact information to the first group of contact information.

5. A computer-implemented method of fraud protection, comprising:
obtaining an image of a physical document, wherein the document comprises at least one of a written communication or a printed communication;
sending the image of the document to a server, wherein the server is configured to:
extract at least one of graphics and text from the image of the document, the text being extracted using optical character recognition;
compare the extracted text to a database of keywords, wherein each keyword has an associated likelihood of fraud determined via a machine learning algorithm trained on instances of known fraud obtained from at least one database maintained by at least one of a financial institution, a government agency, or a law enforcement agency;
determine at least one word from the extracted text that matches a keyword in the database;
calculate, by the machine learning algorithm, a confidence level of fraud for the physical document based on the likelihood of fraud of the at least one determined word that matches and a frequency of appearance in the image of the at least one determined word that matches; and
in response to calculating a confidence level of fraud higher than a pre-defined threshold, communicate an indication to display a notification of the confidence level of fraud through a user interface of a user device associated with a user;

receiving the indication of the confidence level of fraud from the server; and
presenting for display, the notification in the user interface.

6. The method of claim 5 , wherein the image of the document is obtained from an image capturing device associated with a user device.

7. The method of claim 5 , wherein obtaining the image of the document comprises receiving an image file from a user.

8. The method of claim 6 , wherein the user device comprises at least one of a mobile device, a computer, a home appliance, or a home assistant device.

9. The method of claim 5 , wherein the image is obtained from an electronic message extracted from an electronic messaging account associated with a user.

10. The method of claim 5 , wherein the threshold value is between 50% and 100%.

11. A system of fraud protection, comprising:
a memory; and
a processor circuit coupled to the memory and configured to execute instructions causing the processor circuit to:
scrape an image of a physical piece of mail in association with actual mail delivery from a user's electronic messaging account;
extract at least one of text and graphics from the image, the text being extracted using optical character recognition;
compare the extracted text to a database of keywords, wherein each keyword has an associated likelihood of fraud determined via a machine learning algorithm trained on instances of known fraud obtained from at least one database maintained by at least one of a financial institution, a government agency, or a law enforcement agency;
determine at least one word from the extracted text that matches a keyword in the database;
calculate, by the machine learning algorithm, a confidence level of fraud for the physical piece of mail based on the likelihood of fraud of the at least one determined word that matches and a frequency of appearance in the image of the at least one determined word that matches; and
in response to calculating a confidence level of fraud higher than a pre-defined threshold, communicate an indication to display a notification of the confidence level of fraud through a user interface of a user device.

12. The system of claim 11 , wherein the instructions further cause the processor circuit to:
receive one or more reports of potential fraud from one or more users; and
update the machine learning algorithm based on the received one or more reports.

13. The system of claim 11 , wherein the extracted graphics comprise at least one of a logo, a trademark, or a signature.

Assignees

• Capital One Services Llc

Inventors

• Mossoba Michael
• Edwards Joshua
• Cunningham Sarah
• Rothbaum Mark

Classifications

• G06V30/10

  Character recognition · CPC title

• G06V30/413

  Classification of content, e.g. text, photographs or tables · CPC title

• G06F16/90335

  Query processing · CPC title

• G06Q30/0185 Primary

  Product, service or business identity fraud · CPC title

• G06K2209/01

  Physics · mapped topic