FRAUD DETECTION SYSTEM ON DIGITAL CHANNELS

Abstract

This invention relates to a system (1) developed to enable the detection of possible frauds occurring in digital channels by processing and analyzing the past transaction data of bank customers who conduct transactions in digital channels using machine learning techniques.

Description

1 TARIFF FRAUD DETECTION SYSTEM ON DIGITAL CHANNELS Technical Area This invention analyzes the transaction history of bank customers who conduct transactions through digital channels. with the processes of data processing and analysis using machine learning techniques a system that enables the detection of potential fraud occurring through digital channels It is related to the system. Previous Technique Bank customers who conduct transactions through digital channels are vulnerable to fraud. There are many difficulties in determining whether or not there is any left. First of all, the fraudsters Because methods are constantly changing, existing rules and models quickly become outdated. Customer behavior can be lost, which makes anomaly detection more difficult. It varies from person to person, determining whether an unusual transaction is fraud or not. This makes it difficult to discern whether it is a normal customer preference. Also... The very high real-time transaction volume ensures that analyses are performed without delay and without errors. This makes it technically challenging to implement. 20 false positive alarms While this increase negatively impacts customer experience, false negatives are a form of fraud. This can lead to cases being overlooked. In addition to all this, data... Privacy and regulatory restrictions make it difficult to integrate and effectively process data from diverse sources. This stands out as a significant factor limiting its use in this way. Therefore, bank customers who conduct transactions through digital channels should have their transaction history checked. with the processes of data processing and analysis using machine learning techniques 2 a system that enables the detection of potential fraud occurring through digital channels It is understood that the system is needed. A Chinese patent document, CN108510279, which is included in the prior art, 5 methods and systems for identifying bank fraud transactions The method is described as follows: receiving a user's transaction request, processing the transaction request... obtaining customer information and transaction address information, here transaction customer information, includes transaction initiator information and transaction counterparty information; transaction customer information, transaction address information and previously obtained fraud suspect list information It includes steps to detect a fraud based on a crime. The method, a crime 10 The bank can lock the suspect's transaction time and transaction position and take a related action. by the transaction position and the customer's actual position while performing the transaction Fraud prevention and control based on real-time information obtained. This reduces the risk of account theft transactions even further. Brief Description of the Invention The purpose of this invention is to analyze the past transactions of bank customers who conduct transactions on digital channels. Processes of processing transaction data and analyzing it using machine learning techniques. to enable the detection of potential fraud occurring through digital channels 20 The goal is to implement a system developed for this purpose. Detailed Description of the Invention Developed to achieve the purpose of this invention, “Fraud Detection in Digital Channels 25 The "System" is shown in the attached figure; Figure 1 shows a schematic view of the system that is the subject of the invention. 3 The parts shown in the figure are individually numbered, and the corresponding numbers correspond to these numbers. It is given below. 1. System 5 2. Database 3. Server The past transaction data of bank customers who conduct transactions through digital channels Digital 10 through processes of processing and analysis using machine learning techniques In order to enable the detection of potential fraud occurring on the channels The developed invention subject system (1); -to be used in the evaluation of transactions carried out on digital channels to ensure that customer data and transaction data are kept on record at least one structured database (2), 15 -Automatically triggered for transactions carried out on digital channels. obtaining relevant data from the customer database (2), machine learning processing, analysis using techniques and verifying whether customer transactions are genuine at least structured to ensure that its absence is detected and recorded. It includes a server (3). 20 The database (2) in the system (1) which is the subject of the invention, any communication protocol to communicate with the server (3) and exchange data using It is structured. The database (2) shows how many different people the customer transferred to in the last six months. The categorized version of the information about what the customer did is 25 in the last month. The ratio of the number of EFT transactions to the total number of financial transactions, the total number of money transfers made by the customer in the last three months ratio to the number of financial transactions, customer's education level, customer's last six months 4 the number of QR transactions made within it compared to the total number of financial transactions ratio, the ratio of a customer's average debt over the last month to their average debt over the last three months, The total number of transactions in demand deposits for investment purposes in the last ninety days. the ratio of the number of debts to the customer's average debt over the last three months compared to their debt over the last six months The ratio of the number of internet activities in the last ninety days to the average of the total number of activities is 5. ratio of the number of entries to the number of digital channels in the last two days, customer data in the last six months transaction tracking indicator, average number of regularly received EFTs over the last three hundred and sixty days, The three hundred and sixty-day average of the number of demand deposit transactions, the last three hundred and sixty how many times the customer logged into digital banking channels during the day, in the last two days Number of password or passcode changes included, customer's assets in the last six months 10 the ratio of the average debt to the average debt over the last six months, over the last ninety days The number of ATM entries is the working size of the total assets over the last three hundred and sixty days. ratio, ratio of FX assets to total assets over the last three hundred and sixty days, client's age, Number of password or passcode changes in the last week, compared to the customer's last month The ratio of the number of other transactions it has carried out within the last 15 years to its total financial transactions. Number of internet login method changes within two days, customer's data in the last month The ratio of average assets to average liabilities over the last month, the customer's average debt over the last month the number of payment transactions made within it compared to the total number of financial transactions rate, number of SMS password errors in the last two days, customer's last day how many different people the transfer was made to, digital login in the last day 20 the number of money transfer transactions the customer has made in the last three months the ratio of the number of transactions to the total number of financial transactions for money transfer purposes in the last ninety days The ratio of the number of demand deposit transactions to the total number of transactions, within the last three days. Number of address changes, number of updated activity listings in the last two days, last three hundred The average of the number of transactions on the days the product was active within sixty days, the current 25 The oldest of the open accounts is the one whose age in days is the one where the product was last actively used. Number of days elapsed since the day it was used, address change in the last two weeks number, customer's average assets over the last twelve months, debt over the last twelve months ratio to the average, number of address changes in the last day, number of address changes in the last two days Number of redirects to government website, password change in the last week If it happened, information on how many days ago it happened, and digital data from the last week. Channel entry count, maximum delay days for personal loans in the last three months number, number of income and expense views in the last two days, working method, customer's 5 The ratio of the average total debt over the last six months to the average total debt over the last twelve months, The number of transactions in the digital banking channel in the last ninety days is the total number of transactions. the ratio of the number of transactions in demand deposits for payment purposes in the last ninety days the ratio of the total number of transactions to the customer's average activity over the last three months ratio of monthly debt to average, number of new accounts opened in the last ninety days, last 10 The ratio of mobile movements to total movements in ninety days, compared to the last twelve months The maximum delay for a single account is the number of delay days, totaling the delay in the last week. Number of successful additional approvals issued, number of entries in the last three days, number of entries in the last ninety days The ratio of the number of transactions in branches to the total number of transactions, the customer's last six months The number of other transfer transactions it has made within the total financial transactions is 15. ratio of demand deposits used for depositing and / or withdrawing cash in the last ninety days The ratio of the number of transactions to the total number of transactions the customer has made in the last month. The ratio of the number of international transfer transactions to the total number of financial transactions, as of last year The average number of consistently outgoing EFTs over three hundred and sixty days is the highest in the last twelve months. The number of days of delay, the maximum maturity within accounts related to asset products is 20. The product's maturity date in days, the incoming remittance for the last three months and / or Trend over the last twelve months, number of password changes in the last 24 hours, number of password changes in the last two days Number of cash flow images, maximum delay in the last twelve months, last two days Number of asset information views, transaction signing log from the last day. number, number of deposit increases made by the customer in the last week, last 25 The ratio of the number of transactions at ATMs over ninety days to the total number of transactions, over the last two days including the number of tokenization activations and validations in the last three hundred and sixty days ATM transaction count, number of untokenized rooted devices in the last day, customer's last 6 The ratio of average assets in the past month to average assets in the last three months, customer relationship number of days, number of unique open accounts held within the last twelve months, customer's last The ratio of average assets over the last six months to average assets over the last twelve months, effective Information on whether an account exists, the number of foreign currency sales in the last two days, the last two days Number of rejected transactions, number of digital channel logins in the last three months, last 5 The maximum number of days overdue for a credit card in three months is the customer's last three months. the ratio of average assets this month to average assets over the last six months, last two days number of incorrect online password entries, personal loan amounts in the last twelve months maximum delay, number of times card PINs have been set in the last two days, last day Number of online application errors or mobile phone errors within the last thirty days: 10 Number of new accounts opened, number of limits defined in the last two days, last three days including email change count, SIM card change count in the last day, last one Target number of deletions from the accumulation box during the day, limit within the last week. Number of definitions, number of changes to term deposit account renewals in the last day, The number of other transfer transactions made by the customer in the last month is 15 Ratio of total financial transactions to internet and mobile device transactions in the last week. Matching number refers to the foreign currency transfers the customer has made in the last six months. the ratio of the number of transactions to the total number of financial transactions, current maximum delay, number of password changes in the last two days, phone calls in the last week If a change occurred, information on how many days ago it happened, within the last week (20 days). If an email address change occurred, information on how many days ago it happened (within the last two days) is required. This includes the number of SIM card changes and the number of account renewal changes within the last two days. If there was a SIM change within the last week, the days elapsed since the SIM change. The number of days before the address change, if the change of address occurred within the last week. The information indicates that the number of online or mobile phone errors that occurred in the last two days is 25. Number of withdrawals from the term deposit account in the last two days, number of internet withdrawals in the last day and the number of mobile device pairings, the maximum latency of a single account in the last twelve months, last Number of email changes in a single day, phone number changes in the last day 7 customer data in the form of number of changes and number of SIM card changes in the last week. to ensure that information is obtained from relevant sources and kept on record It is being structured. The server (3) in the system (1) which is the subject of the invention, can use any communication protocol 5 to communicate with the database (2) and exchange data using It is structured. The server (3) processes data with a supervised machine learning algorithm. (3) Changes in the volume and number of financial transactions recorded at the base rates, segment information, credibility, critical information change information, digital variables such as the number of operations and images performed on the channels 10 using and analyzing the weights in the supervised machine learning model and to enable the detection of digital channel fraud It is structured. The server (3) uses a classification algorithm for different variables. conducting analyses of what happened on digital channels in previous years The selection of fraudulent transactions as the target variable, fraud (scam) 15 Generating variables to assess the likelihood of being exposed to fraud from various sources and It is configured to enable analysis. Server (3), XGBoost to enable the detection of potential fraud on digital channels with its algorithm is configured. The server (3) is configured to configure the parameters used in the data processing process. expansion, use of different machine learning algorithms and model tracking 20 their studies to detect potential fraud occurring on digital channels It is configured to provide the server (3) to the machine learning model dataset. real-time, instantaneous additional information is fed into the interbank fraud dataset. By developing the target variable set through sharing methods and on digital channels. 25 to ensure the detection of potential frauds that have occurred It is being structured. Industrial Applicability 8 Banks that conduct transactions on digital channels thanks to the system (1) that is the subject of the invention processing customers' past transaction data and machine learning techniques the processes of analyzing and detecting potential fraud occurring through digital channels This is ensured. 5 Based on these fundamental concepts, the invention focuses on "Fraud Detection in Digital Channels". It is possible to develop a wide variety of applications related to the System (1)”, and the invention This cannot be limited to the examples described here, but is primarily stated in the claims. It is like this. 10

Claims

9 REQUESTS 1. Transaction history of bank customers who conduct transactions through digital channels. processes of data processing and analysis using machine learning techniques 5. Detecting potential fraud occurring through digital channels providing; -in the evaluation of transactions carried out on digital channels keeping records of customer data and transaction data to be used containing at least one database (2) configured to provide and -Automatically triggered for transactions carried out on digital channels 10 obtaining the relevant data from the customer database (2), machine processing, analyzing and interpreting customer transactions using learning techniques to ensure that its authenticity is determined and recorded. a system characterized by having at least one server (3) configured (1).

2. To communicate with the server (3) using any communication protocol and with the database (2) configured to carry out data exchange A system like the one in Claim 1, characterized (1).

3. Information on how many different people the customer transferred money to in the last six months. 20 The categorized data shows the customer's EFT transactions over the past month. The ratio of the number of transactions to the total number of financial transactions of the customer in the last three months the number of wire transfers made within the total financial transactions ratio of the number, customer's education level, customer's past six months The ratio of the number of QR transactions to the total number of financial transactions is 25. the customer's average debt over the last month compared to the average debt over the last three months the ratio of the number of investment-related demand deposit transactions in the last ninety days the ratio of the total number of transactions to the customer's average debt over the last three months ratio of debt to average over six months, internet activity in the last ninety days the ratio of the number of movements to the total number of movements on digital channels in the last two days Number of entries, customer activity tracking indicator for the last six months, last three hundred and sixty The average number of regular daily EFT incoming transactions is three times the number of demand deposit transactions. The 160-day average represents the customer's digital 5 experience over the last 360 days. how many times they logged into banking channels, password or other data loss in the last two days Number of password retrieval changes, the customer's average asset value over the last six months ratio to the average debt over six months, ATM transactions in the last ninety days. number, ratio of total assets to working size in the last three hundred and sixty days, the ratio of FX assets to total assets over the last three hundred and sixty days, the client's age, 10 The number of password or PIN change requests made within the last week, the customer's last the total financial value of the number of other transactions made within a month Processing rate, number of internet login method changes in the last two days, the customer's average assets over the last month compared to their average liabilities over the last month the rate is 15% of the customer's payment transactions in the last month. The ratio of the number of short messages to the total number of financial transactions in the last two days. Password error count, how many different people the customer transferred to in the last day the number of digital logins made in the last day, the customer's last three months the number of money transfer transactions it has made within its total financial The ratio to the number of transactions, demand deposits used for money transfers in the last ninety days is 20. the ratio of the number of movements to the total number of movements, address within the last three days number of changes, number of updated activity listings in the last two days, last three the number of transactions on the days the product was active within one hundred and sixty days the average age, in days, of the oldest of the existing open accounts, The number of days that have passed since the last active use of the product, the last two 25-day periods. Number of address changes during the week, customer's assets in the last twelve months the ratio of the average debt to the average debt over the last twelve months, within the last day Number of address changes to the government website within the last two days 11 Number of redirects, how many if a password change occurred in the last week. Information that it happened days ago, digital channel entry within the last week number of days overdue on personal loans in the last three months number, number of income and expense views in the last two days, working method, The customer's average total debt over six months is 5 times the average debt over the last twelve months. the ratio of total activity in the digital banking channel over the last ninety days the ratio of the number of transactions to the total number of transactions, for payment purposes in the last ninety days The ratio of the number of transactions in demand deposit accounts to the total number of transactions, the customer's The ratio of average assets to average liabilities over the last three months, Number of new accounts opened within ninety days, mobile activity in the last ninety days 10 the ratio of the number of transactions to the total number of transactions, single account maximum in the last twelve months the number of days of delay, the total number of days left in the last week Number of successful additional approvals, number of entries in the last three days, number of entries in the last ninety days The ratio of the number of transactions in branches to the total number of transactions, the customer's last six The total number of other transfer transactions he made during the month is 15. Ratio to the number of financial transactions, cash deposits and / or withdrawals in the last ninety days The ratio of the number of transactions in non-interest-based deposits to the total number of transactions, international transfer transactions made by the customer in the last month The ratio of the number of transactions to the total number of financial transactions has been consistently high over the last three hundred and sixty days. The average number of outgoing EFTs has a delay of 20, the highest delay in the last twelve months. number of days, accounts with the maximum maturity value within asset product accounts The product's maturity date in days, the incoming remittance in the last three months and / or the last ten months two-month trend, number of password changes in the last day, last two days number of cash flow statements included, maximum delay in the last twelve months, Number of asset information views in the last two days, 25 in the last day. Transaction signing log count, of transactions made by the customer in the last week. the number of deposit increases, the total number of transactions at ATMs in the last ninety days the ratio of the number of transactions, tokenization activation in the last two days and 12 validation count, number of ATM transactions in the last three hundred and sixty days, number of transactions in the last day Number of untokenized rooted devices, customer's assets in the last month the ratio of the average asset value to the average asset value over the last three months, customer relationship day the number, the number of unique open accounts the customer has had in the last twelve months The ratio of average assets over the last six months to average assets over the last twelve months is 5. Information on whether an active account exists, the number of foreign currency sales in the last two days, Number of rejected transactions in the last two days, digital transactions in the last three months Number of channel logins, maximum credit card late payment overdue in the last three months Number of days delayed, the customer's asset average for the last six months The ratio of assets to the monthly average, incorrect online passwords in the last two days: 10 Number of entries, maximum overdue personal loan in the last twelve months, last two days Number of card PINs set, online application within the last day Number of phone or mobile phone errors, number of new accounts opened in the last thirty days, Number of limit definitions in the last two days, email in the last three days Number of changes, number of SIM card changes in the last day, last day 15 Accumulation box target deletion count, limit within the last week Number of definitions, number of changes to term deposit account renewals in the last day, other transfer transactions made by the customer in the last month The ratio of the number of transactions to the total number of financial transactions on the internet in the last week. and the number of mobile device pairings, which the customer has done in the last six months (20). the number of foreign currency transfer transactions relative to the total number of financial transactions rate, current maximum latency, password change within the last two days number, if the phone was changed within the last week, how many days ago The information that this happened indicates an email change within the last week. If it happened, information on how many days ago it happened, within the last two days SIM 25 number of changes, number of account renewal changes in the last two days, last one If there was a SIM change during the week, the number of days following the SIM change The number of days before the address change, if the change of address occurred within the last week. 13 The information that it happened was obtained via online or mobile phone within the last two days. number of errors, number of withdrawals from the term deposit account in the last two days, last one Number of internet and mobile device pairings during the day, single account in the last twelve months. maximum delay, number of email changes in the last day, last one Number of phone number changes during the day and SIM 5 in the last week Obtaining customer information in the form of number of changes from relevant sources and the database structured to ensure that it is kept on record (2) A system like the one in Claim 1 or 2, characterized (1).

4. Communicate with the database (2) using any communication protocol 10 and characterized by the server (3) configured to carry out data exchange. a system like any of the above-mentioned requests (1).

5. With the supervised machine learning algorithm, records in the database (3) Rates of change in the volume and number of financial transactions recorded, segment 15 information, credibility, critical information change information, digital channels variables such as the number of operations performed and the number of images viewed analysis of weights in the supervised machine learning model using to ensure that and digital channel fraud is detected Any of the above requests characterized by the configured server (3) 20 a system like one of them (1).

6. Classification algorithm for analysis of different variables. the implementation of which took place in digital channels in previous years The selection of fraudulent transactions as the target variable, fraud 25 variables related to the likelihood of being exposed to (scam) fraud with the server (3) configured to enable the production and analysis a system like any of the above characterized demands (1). 14 7. Detecting potential fraud on digital channels with the XGBoost algorithm. characterized by the server (3) configured to enable it to be done a system like any of the above requests (1).

8. Expanding the parameters used in the data processing process, different machines 5 digital learning algorithms are used and model tracking studies are conducted. to enable the detection of potential fraud occurring on the channels from the above requests characterized by the server (3) configured for a system like any other (1).

9. Adding real-time, instantaneous information to the machine learning model dataset. through methods of feeding and sharing interbank fraud data sets by developing a set of target variables, potential occurrences on digital channels Server configured to enable the detection of fraud (3) A system like any of the above-mentioned systems characterized by 15 (1).

Images