Selective authentication based on similarities of ecommerce transactions from a same user terminal across financial accounts

Patent metadata

Field	Value
Publication number	US-2016203485-A1
Application number	US-201514592136-A
Country	US
Kind code	A1
Filing date	Jan 8, 2015
Priority date	Jan 8, 2015
Publication date	Jul 14, 2016
Grant date	—

Abstract

A method of operating a computer system includes receiving from a merchant node an eCommerce authentication request for a pending eCommerce transaction associated with a user terminal. The eCommerce authentication request contains transaction information of the pending eCommerce transaction that includes a user terminal identifier. A risk score for the pending eCommerce transaction is generated based on similarities between the transaction information of the pending eCommerce transaction and a cluster of historical eCommerce transactions each containing transaction information including a user terminal identifier that matches the user terminal identifier of the pending eCommerce transaction. The eCommerce authentication request is selectively provided to an authentication node based on the risk score.

First claim

1 . A method of operating a computer system comprising:
receiving from a merchant node an eCommerce authentication request for a pending eCommerce transaction associated with a user terminal, the eCommerce authentication request containing transaction information of the pending eCommerce transaction that comprises a user terminal identifier; generating a risk score for the pending eCommerce transaction based on similarities between the transaction information of the pending eCommerce transaction and a cluster of historical eCommerce transactions each containing transaction information comprising a user terminal identifier matching the user terminal identifier of the pending eCommerce transaction; and selectively providing the eCommerce authentication request to an authentication node based on the risk score.

2 . The method of claim 1 , further comprising:
storing in a repository the historical eCommerce transactions received from finance issuer nodes, each of the historical eCommerce transactions containing a user terminal identifier, a financial account number, and an indications of whether fraud was detected; generating user terminal groupings of the historical eCommerce transactions in the repository based on the user terminal identifiers, each of the historical eCommerce transactions in any one of the user terminal groups having user terminal identifiers that match; and training a non-linear analytical model with the user terminal groupings of the historical eCommerce transactions, wherein the generating a risk score comprises:
processing the transaction information of the pending eCommerce transaction through the non-linear analytical model to generate the risk score.

3 . The method of claim 2 , wherein the training the non-linear analytical model with the user terminal groupings of the historical eCommerce transactions, comprises.
training a neural network model based on the user terminal identifiers, the financial account number, and the indications of whether fraud was detected for each of the user terminal groupings of the historical eCommerce transactions.

4 . The method of claim 3 ,
wherein the neural network model comprises an input layer comprising input nodes, a sequence of neural network layers each comprising a plurality of weight nodes, and an output layer comprising an output node; the method further comprising:
operating the input nodes of the input layer to each receive different content of the transaction information of the pending eCommerce transaction and output a value;
operating the weight nodes of a first one of the sequence of neural network layers using weight values to mathematically combine values that are output by the input nodes to generate combined values;
operating the weight nodes of a last one of the sequence of neural network layers using weight values to mathematically combine the combined values from a plurality of weight nodes of a previous one of the sequence of neural network layers to generate combined values; and
operating the output node of the output layer to combine the combined values from the weight nodes of the last one of the sequence of neural network layers to generate the risk score.

5 . The method of claim 4 , wherein the training the neural network model based on the user terminal identifiers, the financial account numbers, and the indications of whether fraud was detected for each of the user terminal groupings of the historical eCommerce transactions, comprises
training different groupings of the weight values of at least one of the neural network layers based on different corresponding ones of the user terminal groupings of the user terminal identifiers, the financial account numbers, and the indications of whether fraud was detected for the historical eCommerce transactions.

6 . The method of claim 1 , further comprising:
storing in a repository the historical eCommerce transactions from the authentication node, each of the historical eCommerce transactions containing a user terminal identifier, a financial account number, and an indication of whether an associated historical eCommerce authentication request passed authentication by the authentication node; generating user terminal groupings of the historical eCommerce transactions in the repository based on the user terminal identifiers, each of the historical eCommerce transactions in any one of the user terminal groups having user terminal identifiers that match; and training a non-linear analytical model with the user terminal groupings of the historical eCommerce transactions, wherein the generating a risk score comprises:
processing the transaction information of the pending eCommerce transaction through the non-linear analytical model to generate the risk score.

7 . The method of claim 6 , wherein the training the non-linear analytical model with the user terminal groupings of the historical eCommerce transactions, comprises.
training a neural network model based on the user terminal identifiers, the financial account numbers, and the indications of whether associated historical eCommerce authentication requests passed authentication by the authentication node for each of the user terminal groupings of the historical eCommerce transactions.

8 . The method of claim 7 ,
wherein the neural network model comprises an input layer comprising input nodes, a sequence of neural network layers each comprising a plurality of weight nodes, and an output layer comprising an output node; the method further comprising:
operating the input nodes of the input layer to each receive different content of the transaction information of the pending eCommerce transaction and output a value;
operating the weight nodes of a first one of the sequence of neural network layers using weight values to mathematically combine values that are output by the input nodes to generate combined values;
operating the weight nodes of a last one of the sequence of neural network layers using weight values to mathematically combine the combined values from a plurality of weight nodes of a previous one of the sequence of neural network layers to generate combined values; and
operating the output node of the output layer to combine the combined values from the weight nodes of the last one of the sequence of neural network layers to generate the risk score.

9 . The method of claim 8 , wherein the training the neural network model based on the user terminal identifiers, the financial account numbers, and the indications of whether associated historical eCommerce authentication requests passed authentication by the authentication node for each of the user terminal groupings of the historical eCommerce transactions, comprises
training different groupings of the weight values of at least one of the neural network layers based on different corresponding ones of the user terminal groupings of the user terminal identifiers, the financial account numbers, and the indications of whether associated historical eCommerce authentication requests passed authentication by the authentication node.

10 . The method of claim 1 , wherein:
the user terminal identifier comprises a network address of a user terminal that is a source of the pending eCommerce transaction; and the risk score for the pending eCommerce transaction is generated based on similarities between the transaction information of the pending eCommerce transaction and the cluster of historical eCommerce transactions each containing transaction information comprising a network address for a user terminal that was the source of the historical eCommerce transaction which matches the network address of the user terminal that is a source of the pending eCommerc

Assignees

• Ca Inc

Inventors

• Subramanian Revathi
• Dulany Paul C
• Gong Hongrui
• Shah Kannan Shashank

Classifications

• G06Q20/405

  Establishing or using transaction specific rules · CPC title

• G06Q20/4016 Primary

  involving fraud or risk level assessment in transaction processing · CPC title

• G06Q30/00

  Commerce · CPC title