Self-labelling of fraud risk in a transaction processing system
AssigneeCheckmate Network Ltd.
InventorsIsrael Hanania Rosilio, Aviel Meir Erdis
Life of this patent
Estimating the term…
Abstract
The presently disclosed subject matter relates to detection and mitigation of financial fraud, and in particular to implementation of systems for training machine learning models in such systems. It involves self-labelling of fraud risk in a transaction processing system.
Claims7
A processing circuitry-based method of assigning machine learning training labels indicative of a level of risk, to accounts of a financial institution, the method comprising:
- a) identifying a first group of accounts of the financial institution as suspected of fraudulent behavior, based on degrees of matching between:
- i. account history data of each respective account, and
- ii. one or more given fraud patterns;
- b) identifying a second suspected group of accounts, the identifying comprising:
- i. for one or more accounts of the first suspected group of accounts:
- determining respective degrees of personally identifiable information (PII) similarity between the respective account of the first suspected group of accounts, and other accounts of the first suspected group of accounts, and
- ii. selecting accounts for which the respective determined PII similarity meets a threshold degree of PII similarity;
- c) identifying a third suspected group of accounts, the identifying comprising:
- i. for one or more accounts of the second suspected group of accounts:
- determining respective degrees of financial transaction activity between the respective account and other accounts of the second suspected group, and
- ii. selecting accounts for which the respective degree of financial transaction activity meets a threshold degree of financial activity;
- d) assigning a machine learning training label indicative of fraud risk to one or more accounts of the third suspected groups of accounts.
A system of assigning machine learning training labels indicative of a level of risk, to accounts of a financial institution, the system comprising a processing circuitry configured to:
- a) identify a first group of accounts of the financial institution as suspected of fraudulent behavior, based on degrees of matching between:
- i. account history data of each respective account, and
- ii. one or more given fraud patterns;
- b) identify a second suspected group of accounts, the identifying comprising:
- i. for one or more accounts of the first suspected group of accounts:
- determining respective degrees of personally identifiable information (PII) similarity between the respective account of the first suspected group of accounts, and other accounts of the first suspected group of accounts, and
- ii. selecting accounts for which the respective determined PII similarity meets a threshold degree of PII similarity;
- c) identify a third suspected group of accounts, the identifying comprising:
- i. for one or more accounts of the second suspected group of accounts:
- determining respective degrees of financial transaction activity between the respective account and other accounts of the second suspected group, and
- ii. selecting accounts for which the respective degree of financial transaction activity meets a threshold degree of financial activity;
- d) assign a machine learning training label indicative of fraud risk to one or more accounts of the third suspected groups of accounts.
A computer program product comprising a computer readable non-transitory storage medium containing program instructions, which program instructions when read by a processor, cause the processing circuitry to perform a method of assigning machine learning training labels indicative of a level of risk, to accounts of a financial institution, the method comprising:
- a) identifying a first group of accounts of the financial institution as suspected of fraudulent behavior, based on degrees of matching between:
- i. account history data of each respective account, and
- ii. one or more given fraud patterns;
- b) identifying a second suspected group of accounts, the identifying comprising:
- i. for one or more accounts of the first suspected group of accounts:
- determining respective degrees of personally identifiable information (PII) similarity between the respective account of the first suspected group of accounts, and other accounts of the first suspected group of accounts, and
- ii. selecting accounts for which the respective determined PII similarity meets a threshold degree of PII similarity;
- c) identifying a third suspected group of accounts, the identifying comprising:
- i. for one or more accounts of the second suspected group of accounts:
- determining respective degrees of financial transaction activity between the respective account and other accounts of the second suspected group, and
- ii. selecting accounts for which the respective degree of financial transaction activity meets a threshold degree of financial activity;
- d) assigning a machine learning training label indicative of fraud risk to one or more accounts of the third suspected groups of accounts.
Dependencies are inferred from the claim text.