Granted Patent B1

US 12,731,426 · App. 18/401,603 · Granted Sep 8, 2026

Inventors: Ashwin Pingali (Parker, CO); Arman Mohseni-Kabir (Carlsbad, CA); Marzieh Mehdizadeh (Sunnyvale, CA); Sudha Vijayakumar (Campbell, CA); Lulu Li (Millbrae, CA); Osvaldo Driollet (Carlsbad, CA)

Assignee: Jumio Corporation

G06V30/41

Abstract

Claims (72)

1 . A computer-implemented method, comprising:

receiving a first image of a document associated with a transaction;

generating a fraud risk score associated with the transaction based on a verification process on the first image of the document;

storing the fraud risk score in association with the transaction in a database;

receiving additional images associated with the document;

processing the additional images according to one or more pattern recognition functions;

evaluating the fraud risk score associated with the transaction based on at least one outcome of the pattern recognition functions;

determining a new fraud risk score associated with the transaction; and

storing the new fraud risk score as the fraud risk score in association with the transaction in the database;

identifying, for each image received, a plurality of data fragments associated with the transaction, each data fragment comprising identifying information about a person in the transaction;

generating a transaction graph based on received images and storing the transaction graph in the database;

determining one or more patterns of transactional data in the database, wherein the one or more patterns are stored in a knowledge graph in the database; and

determining one or more meta-patterns from the one or more patterns stored in the knowledge graph in the database, wherein the one or more meta-patterns are determined based on one or more fraudulent transactions in the one or more patterns.

2 . The method of claim 1 , wherein the plurality of data fragments comprises photo fragments, context fragments, channel fragments, and identity fragments, and generating the transaction graph comprises enforcing a uniqueness property on identity fragment nodes such that identical identity fragments across transactions share a single fragment node.

3 . The method of claim 1 , further comprising:

for each data fragment about the person in the transaction, generating a unique identifier as a node in a knowledge graph stored in the database;

generating a cluster based on each data fragment; and

storing the cluster as a node in the database.

4 . The method of claim 3 , further comprising populating the knowledge graph with probabilistic knowledge data based on clusters of transactional data in the database.

5 . The method of claim 2 , wherein the plurality of data fragments includes photo fragments, context fragments, channel fragments, and identity fragments.

6 . The method of claim 2 , wherein determining the one or more patterns of transation data comprises identifying a pattern wherein one document identity fragment connects multiple person identity fragments.

7 . The method of claim 2 , wherein the knowledge graph comprises multi-dimensional data.

8 . The method of claim 6 , wherein the one or more meta-patterns comprise a mixed pattern connecting multiple topology patterns through shared den fraudulent transactions.

9 . The method of claim 3 , further comprising:

creating fingerprints of one or more sub-graphs of the knowledge graph;

converting the fingerprints into vectors, comparing the vectors for similarity; and

identifying a sub-graph having a low similarity as a discrepancy and determining a risk associated with the discrepancy,

wherein the risk is generated as a probabilistic data value based on a cluster hierarchy.

10 . A method of generating a plurality of knowledge graphs, comprising:

generating a knowledge graph comprising a plurality of identity fragments and a plurality of associations derived from received transactions encoded as nodes in a transactional graph;

generating a projection of the knowledge graph to identify one or more patterns;

transforming the one or more patterns into one or more meta-patterns;

clustering the nodes in the knowledge graph based on an identity fragment;

generating a fraud risk score associated with each cluster using a pattern predictive index;

storing the fraud risk score associated with each cluster as a data fragment node in the knowledge graph in a database;

identifying a subgraph of the nodes in the knowledge graph based on one or more known typology patterns; and

storing the subgraph of the nodes as a pattern node in the knowledge graph in the database.

11 . The method of claim 10 , wherein the storing occurs in a cloud-based data storage system.

12 . The method of claim 10 , wherein a connection between two of the nodes in the knowledge graph is generated based on encoded information associated with the received transactions.

13 . The method of claim 10 , wherein transforming the one or more patterns into one or more meta-patterns comprises condensing each pattern into a pattern node connected to multiple typology pattern nodes through shared fraudulent transaction nodes.

14 . A system comprising:

a processor; and

a memory, the memory storing instructions that, when executed by the processor, cause the system to:

generate a knowledge graph comprising a plurality of identity fragments and a plurality of associations derived from received transactions encoded as nodes in a transactional graph;

generate a projection of the knowledge graph to identify one or more patterns;

transform the one or more patterns into one or more meta-patterns;

cluster the nodes in the knowledge graph based on an identity fragment;

generate a fraud risk score associated with each cluster using a pattern predictive index;

store the fraud risk score associated with each cluster as a data fragment node in the knowledge graph in a database;

identify a subgraph of the nodes in the knowledge graph based on one or more known typology patterns; and

store the subgraph of the nodes as a pattern node in the knowledge graph in the database.

15 . The system of claim 14 , wherein the storing occurs in a cloud-based data storage system.

16 . The system of claim 14 , wherein a connection between two of the nodes in the knowledge graph is generated based on encoded information associated with the received transactions.

17 . The system of claim 14 , wherein the memory includes further instructions that, when executed by the processor, cause the system to:

transform the one or more patterns into one or more meta-patterns, wherein the transform comprises condensing each pattern into a pattern node connected to multiple typology pattern nodes through shared fraudulent transaction nodes.

18 . The system of claim 14 , wherein the memory includes further instructions that, when executed by the processor, cause the system to:

receive a first image of a document associated with a transaction;

generate a fraud risk score associated with the transaction based on a verification process on the first image of the document;

store the fraud risk score in association with the transaction in a database;

receive additional images associated with the document;

process the additional images according to one or more pattern recognition functions;

evaluate the fraud risk score associated with the transaction based on at least one outcome of the pattern recognition functions;

determine a new fraud risk score associated with the transaction; and

store the new fraud risk score as the fraud risk score in association with the transaction in the database.

19 . The system of claim 14 , wherein the memory includes further instructions that, when executed by the processor, cause the system to:

for each image received, identify a plurality of data fragments associated with a transaction, each data fragment comprising identifying information about a person in the transaction;

generate a transaction graph based on received images; and

store the transaction graph in the database.

20 . The system of claim 14 , wherein the memory includes further instructions that, when executed by the processor, cause the system to:

for each data fragment about a person in a transaction received, generate a unique identifier in the knowledge graph stored in the database;

generate a cluster based on each data fragment; and

store the cluster as a node in the knowledge graph in the database.

Assignments (1)

References Cited (34)