System and method for detecting internet fraud using a machine learning model
AssigneeSource Ltd.
InventorsShmuel UR, Ilya DUBINSKY
Life of this patent
Estimating the term…
Abstract
A system and method for detecting fraud may scan (e.g. by computer server) a target website provided by another server to identify products provided by the website. Based on the scanning, a process may calculate probabilities for combinations of the identified products, where a cost for each combination equals a cost of a new transaction to take place at the target website, and generate an authentication score for the transaction based on the calculated probabilities.
Claims20
A computerized method of detecting internet fraud, the method comprising:
- determining, by a computer processor, one or more combinations of one or more items, wherein a cost for each of the determined one or more combinations equals a cost for a transaction to take place at a target website;
- for each of the items in each determined combination, predicting, by a machine learning model, a likelihood value for a quantity of the item in the determined combination, wherein the predicted likelihood value is indicative of whether the quantity of the item in the determined combination is common in historical data, wherein the predicting of the likelihood value comprises inputting a tag describing the item into the machine learning model;
- generating, by the processor, a probability of fraud for the transaction to take place at the target website based on one or more of the predicted likelihood values;
- and
- blocking or enabling the transaction to take place at the target website based on the generated probability of fraud.
A computerized system for detecting internet fraud, the system comprising:
- a memory;
- and
- one or more processors configured to:
- determine one or more combinations of one or more items, wherein a cost for each of the determined one or more combinations equals a cost for a transaction to take place at a target website;
- for each of the items in each determined combination, predict, by a machine learning model, a likelihood value for a quantity of the item in the determined combination, wherein the predicted likelihood value is indicative of whether the quantity of the item in the determined combination is common in historical data, wherein the predicting of the likelihood value comprises inputting a tag describing the item into the machine learning model;
- generate a probability of fraud for the transaction to take place at the target website based on one or more of the predicted likelihood values;
- and
- block or enable the transaction to take place at the target website based on the generated probability of fraud.
A computerized method of detecting internet fraud, the method comprising:
- computing, by a computer processor, one or more subsets of one or more items, wherein a cost for each of the determined one or more subsets equals a cost for a transaction to take place at a website;
- for each of the items in each determined subset, generating, by a machine learning model, a probability value for a number of units of the item in the determined subset, wherein the generated probability value is indicative of whether the number of units of the item in the determined subset is common in historical data, wherein the generating of the probability value comprises inputting a label describing the item into the machine learning model;
- generating, by the processor, an authentication score for the transaction to take place at the target website based on one or more of the generated probability values;
- and
- preventing or permitting the transaction to take place at the target website based on the generated authentication score.
Dependencies are inferred from the claim text.