Machine learning based detection of fraudulent acquirer transactions

Abstract

Examples described herein relate to apparatuses and methods of detecting fraudulent activity at an automated teller machine (ATM) using a machine learning model. A method includes receiving ATM activity data indicative of one or more withdrawal transactions at one or more ATMs using a transaction card, receiving transaction data and ATM data, ingesting the transaction data and the ATM data, analyzing the ingested transaction data and the ingested ATM data using a machine learning model, determining that the ingested transaction data and the ingested ATM data indicate fraudulent activity using the machine learning model, and performing one or more remedial actions based on the determination of fraudulent activity using the machine learning model.

What is claimed is: 
     
         1 . A method comprising:
 receiving, by a provider computing system, (i) automated teller machine (ATM) activity data indicative of one or more withdrawal transactions at one or more ATMs, the one or more withdrawal transactions performed using a transaction card, and (ii) transaction data and ATM data corresponding to the one or more withdrawal transactions;   generating, by the provider computing system, an indication that the transaction data and the ATM data correspond to fraudulent activity by providing the transaction data and the ATM data as input to a machine learning model, the machine learning model trained to output a type of fraudulent activity based on a set of training data that indicates a number of ATMs and a withdrawal amount involved in known fraudulent activities; and   selecting, by the provider computing system, one or more remedial actions responsive to the determination of fraudulent activity using the machine learning model, the one or more remedial actions selected based on a success rate of previously selected remedial actions for the type of the fraudulent activity detected using the machine learning model,   wherein the one or more remedial actions comprise at least one of canceling the transaction card, providing a notification to a card-issuing entity associated with the transaction card, transmitting a notification to a user device, or capturing an image of a fraudster at the one or more ATMs.   
     
     
         2 . The method of  claim 1 , wherein the machine learning model uses a Decision Tree Classifier Algorithm. 
     
     
         3 . The method of  claim 1 , further comprising analyzing, by the provider computing system, the transaction data and the ATM data, wherein analyzing the transaction data and the ATM data comprises communication between a fraud analysis circuit and an adaptive processing circuit. 
     
     
         4 . The method of  claim 3 , wherein the adaptive processing circuit provides an updated machine learning model to the fraud analysis circuit to use as part of analyzing the transaction data and the ATM data. 
     
     
         5 . The method of  claim 4 , wherein the adaptive processing circuit retrieves stored ATM data, stored transaction data, and an original machine learning model to develop the updated machine learning model. 
     
     
         6 . The method of  claim 5 , wherein the adaptive processing circuit retrieves historical fraud information from one or more databases of the provider computing system and retrains the original machine learning model based on the historical fraud information to develop the updated machine learning model. 
     
     
         7 . The method of  claim 1 , further comprising determining, by the provider computing system, the one or more remedial actions based on the type of fraudulent activity. 
     
     
         8 . The method of  claim 1 , further comprising at least one of canceling the transaction card or notifying the card-issuing entity associated with the transaction card that the transaction data and the ATM data indicate fraudulent activity. 
     
     
         9 . The method of  claim 1 , further comprising training, by the provider computing system, the machine learning model using a set of training data. 
     
     
         10 . The method of  claim 1 , wherein the provider computing system receives at least one of the transaction data or the ATM data from the one or more ATMs. 
     
     
         11 . A provider computing system comprising a network interface and a processing circuit configured to:
 receive (i) ATM activity data indicative of one or more withdrawal transactions at one or more ATMs, the one or more withdrawal transactions performed using a transaction card, and (ii) transaction data and ATM data corresponding to the one or more withdrawal transactions;   generate an indication that the transaction data and the ATM data correspond to fraudulent activity by providing the transaction data and the ATM data as input to a machine learning model, the machine learning model trained to output a type of fraudulent activity based on a set of training data that indicates a number of ATMs and a withdrawal amount involved in known fraudulent activities; and   select one or more remedial actions responsive to the determination of fraudulent activity using the machine learning model, the one or more remedial actions selected based on a success rate of previously selected remedial actions for the type of the fraudulent activity detected using the machine learning model;   wherein the one or more remedial actions comprise at least one of canceling the transaction card, transmitting a notification to a card-issuing entity associated with the transaction card, transmitting a notification to a user device, or capturing an image of a fraudster at the one or more ATMs.   
     
     
         13 . The provider computing system of  claim 11 , wherein the machine learning model uses a Decision Tree Classifier Algorithm. 
     
     
         14 . The provider computing system of  claim 11 , wherein analysis of the transaction data and the ATM data comprises communication between a fraud analysis circuit and an adaptive processing circuit, the adaptive processing circuit providing an updated machine learning model to the fraud analysis circuit to use as part of the analysis of the transaction data and the ATM data. 
     
     
         15 . The provider computing system of  claim 14 , wherein the adaptive processing circuit retrieves stored ATM data from an ATM database, stored transaction data from a transaction database, and an original machine learning model to develop the updated machine learning model. 
     
     
         16 . The provider computing system of  claim 15 , wherein the adaptive processing circuit retrieves historical fraud information from one or more databases of the provider computing system and retrains the original machine learning model based on the historical fraud information to develop the updated machine learning model. 
     
     
         17 . The provider computing system of  claim 11 , the processing circuit further configured to determine the one or more remedial actions based on the type of fraudulent activity. 
     
     
         18 . The provider computing system of  claim 17 , the processing circuit further configured to perform at least one of canceling the transaction card or notifying the card-issuing entity associated with the transaction card based on the determination that the transaction data and the ATM data indicate fraudulent activity. 
     
     
         19 . The provider computing system of  claim 11 , the processing circuit further configured to train the machine learning model using a set of training data. 
     
     
         20 . The provider computing system of  claim 11 , the processing circuit further configured to receive at least one of the transaction data or the ATM data from the one or more ATMs.