# US 2022/0070292 A1 · Systems and methods employing graph-derived features for fraud detection

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2022/0070292 A1
- Document ID: US20220070292A1
- Status: Published application
- Kind: A1
- Application number: 17/201,849
- Filed: 2021-03-15
- Published: 2022-03-03
- Earliest claimed: 2020-03-17
- Assignees: Pindrop Security, Inc.
- Applicants: Pindrop Security, Inc.
- Inventors: Ricardo CASAL; Theo WALKER; Kailash PATIL; John CORNWELL
- Web page: https://patentgrep.com/patent/US20220070292A1

## CPC

- H04M 3/2281: {Call monitoring, e.g. for law enforcement purposes; Call tracing; Detection or prevention of malicious calls}
- G06F 18/214: Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06N 3/09: Supervised learning
- G06N 20/00: Machine learning
- H04M 3/42059: {Making use of the calling party identifier}
- G06N 3/08: Learning methods
- G06N 20/10: using kernel methods, e.g. support vector machines [SVM]
- G06N 20/20: Ensemble learning
- H04M 3/42042: {Notifying the called party of information on the calling party (details within substation equipment H04M1/57, signalling details H04Q3/72)}
- H04M 3/51: Centralised call answering arrangements requiring operator intervention {, e.g. call or contact centers for telemarketing}
- H04M 2203/551: Call history
- H04M 2203/556: Statistical analysis and interpretation
- H04M 2203/6027: Fraud preventions
- H04M: TELEPHONIC COMMUNICATION (circuits for controlling other apparatus via a telephone cable and not involving telephone switching apparatus G08)
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Embodiments described herein provide for performing a risk assessment using graph-derived features of a user interaction. A computer receives interaction information and infers information from the interaction based on information provided to the computer by a communication channel used in transmitting the interaction information. The computer may determine a claimed identity of the user associated with the user interaction. The computer may extract features from the inferred identity and claimed identity. The computer generates a graph representing the structural relationship between the communication channels and claimed identities associated with the inferred identity and claimed identity. The computer may extract additional features from the inferred identity and claimed identity using the graph. The computer may apply the features to a machine learning model to generate a risk score indicating the probability of a fraudulent interaction associated with the user interaction.

## Claims

### Claim 1

A computer-implemented method for assessing a risk of fraud, the method comprising:
obtaining, by a computer, an inferred identity from inbound call data associated with an inbound call and an inbound caller;
obtaining, by the computer, an identity claim associated with the inbound caller;
extracting, by the computer, a first set of features from the inbound call data;
generating, by the computer, a graph structure based upon the inferred identity, the identity claim, a set of prior inferred identities, and a set of prior identity claims associated with the inbound call;
extracting, by the computer, a second set of features from the graph structure; and
applying, by the computer, a machine learning model on the first set of features and the second set of features to generate a risk score for the inbound call.

### Claim 2 (depends on 1)

The method according to claim 1, further comprising applying, by the computer, a classification model on the risk score to determine a risk classification.

### Claim 3 (depends on 1)

The method according to claim 1, further comprising transmitting, by the computer, a risk assessment output to a third-party device, the risk assessment output indicating at least one of the risk score and a risk classification.

### Claim 4 (depends on 1)

The method according to claim 1, further comprising storing, by the computer, a call data record in a record database including the inbound call data, the inferred identity, the identity claim, and the first set of features.

### Claim 5 (depends on 1)

The method according to claim 1, further comprising:
generating, by the computer, a second graph structure based upon a plurality of prior inferred identities and a plurality of prior identity claims associated with a plurality of prior calls, each of the plurality of prior inferred identities associated with the respective prior identity claim and the respective prior call; and
identifying, by the computer, the graph structure embedded in the second graph structure according to the inferred identity and the identity claim associated with the inbound call.

### Claim 6 (depends on 5)

The method according to claim 5, wherein the second graph structure is generated according to a plurality of call data records stored in a record database, the plurality of call data records storing prior call data, prior inferred identities, prior identity claims, and the first set of features associated with prior calls.

### Claim 7 (depends on 1)

The method according to claim 1, wherein the inferred identity is at least one of:
an ANI in a phone channel,
a first IP address associated with a transaction, and
a second IP address associated with voice samples for interactions between Internet of Things (IoT) devices and a provider server.

### Claim 8 (depends on 1)

The method according to claim 1, wherein the second set of features includes features extracted directly from the graph structure.

### Claim 9 (depends on 1)

The method according to claim 1, wherein the second set of features includes similarity based index methods indicating a similarity between a set of inferred identities and a set of identity claims of the graph structure.

### Claim 10 (depends on 9)

The method according to claim 9, further comprising:
determining, by the computer, a similarity score associated with the similarity based index methods by determining at least one of:
a first intersection between a set of neighbors of inferred identities and a set of neighbors of neighbors of identity claims in the graph structure;
a second intersection between a set of neighbors of neighbors of inferred identities and a set of neighbors of identity claims in the graph structure;
a first union between a set of neighbors of inferred identities and a set of neighbors of neighbors of identity claims in the graph structure; and
a second union between a set of neighbors of neighbors of inferred identities and a set of neighbors of identity claims in the graph structure.

### Claim 11

A system comprising:
a database comprising non-transitory memory configured to store prior call data; and
a server comprising a processor configured to:
obtain an inferred identity from inbound call data associated with an inbound call and an inbound caller;
obtain an identity claim associated with the inbound caller;
extract a first set of features from the inbound call data;
generate a graph structure based upon the inferred identity, the identity claim, a set of prior inferred identities, and a set of prior identity claims associated with the inbound call;
extract a second set of features from the graph structure; and
apply a machine leaning model on the first set of features and the second set of features to generate a risk score for the inbound call.

### Claim 12 (depends on 11)

The system according to claim 11, wherein the server is further configured to apply a classification model on the risk score to determine a risk classification.

### Claim 13 (depends on 11)

The system according to claim 11, wherein the server is further configured to transmit a risk assessment output to a third-party device, the risk assessment output indicating at least one of the risk score and a risk classification.

### Claim 14 (depends on 11)

The system according to claim 11, wherein the server is further configured to store a call data record in a record database including the inbound call data, the inferred identity, the identity claim, and the first set of features.

### Claim 15 (depends on 11)

The system according to claim 11, wherein the server is further configured to:
generate a second graph structure based upon a plurality of prior inferred identities and a plurality of prior identity claims associated with a plurality of prior calls, each of the plurality of prior inferred identities associated with the respective prior identity claim and the respective prior call; and
identify the graph structure embedded in the second graph structure according to the inferred identity and the identity claim associated with the inbound call.

### Claim 16 (depends on 15)

The system according to claim 15, wherein the second graph structure is generated according to a plurality of call data records stored in a record database, the plurality of call data records storing prior call data, prior inferred identities, prior identity claims, and the first set of features associated with prior calls.

### Claim 17 (depends on 11)

The system according to claim 11, wherein the inferred identity is at least one of:
an ANI in a phone channel,
a first IP address associated with a transaction, and
a second IP address associated with voice samples for interactions between Internet of Things (IoT) devices and a provider server.

### Claim 18 (depends on 11)

The system according to claim 11, wherein the second set of features includes features extracted directly from the graph structure.

### Claim 19 (depends on 11)

The system according to claim 11, wherein the second set of features includes similarity based index methods indicating a similarity between a set of inferred identities and a set of identity claims of the graph structure.

### Claim 20 (depends on 19)

The system according to claim 19, wherein the server is further configured to:
determine a similarity score associated with the similarity based index methods by determining at least one of:
a first intersection between a set of neighbors of inferred identities and a set of neighbors of neighbors of identity claims in the graph structure;
a second intersection between a set of neighbors of neighbors of inferred identities and a set of neighbors of identity claims in the graph structure;
a first union between a set of neighbors of inferred identities and a set of neighbors of neighbors of identity claims in the graph structure; and
a second union between a set of neighbors of neighbors of inferred identities and a set of neighbors of identity claims in the graph structure.
