How to help banks build anti fraud model


Nowadays, transaction frauds such as credit card swindlers and telecom fraud can always arouse the public’s sensitive nerves and bring about continuous worries about the security of “wallets”.

A lawbreaker swiped 92800 yuan from consumers in a jewelry store with forged credit cards, China reported. After hearing the case, the court ruled that the card issuing bank, the acquiring institution and the merchant should jointly bear the loss. Among them, the defendant’s Bank should compensate 37120 yuan, and the merchant should bear 2784 yuan.

How to help banks build anti fraud model

Transaction fraud has become a black industry chain

This kind of incident has been common. How did credit card swipe happen? Is there any way to block this behavior? To explain this, let’s first take a look at the card transaction process.

In the offline transaction, after the cardholder swipes the card by the merchant, the acquiring bank will first judge whether the card is available; after confirmation, it will indirectly request transaction authorization from the card issuing bank through bank card organizations such as China UnionPay; if the card issuing bank agrees to authorize, the authorization shall be sent back through the bank card organization, and then the card holder and the merchant can conduct the transaction.

How to help banks build anti fraud model

In the whole process of bank card transaction, the issuing bank has the obligation to ensure the card holder’s safety; the acquiring institution is responsible for the management of the business of acquiring business and the security of transaction and information; the cardholder has the obligation to keep the bank card and password properly; the authorized merchant has the obligation to verify the true identity of the cardholder and the authenticity of the bank card; any of these links If there is any problem, it will bring space for the lawbreakers to gain profits through fraud.

The analysis shows that the common fraud modes are application fraud and transaction fraud. The cases mentioned at the beginning of the article belong to counterfeit card fraud, which is the most common fraud means in transaction fraud.

How to help banks build anti fraud model

Analysis shows that criminals often forge bank cards in the following ways:

1. Generate software simulation card number through account.

2. Use the card reader (side recorder) with memory storage device to record the real bank card information, paste magnetic stripe on the blank card, input the code of the side record, and forge the card.

3. To the lost card, stolen card, not reached card (not sent or enabled bank card), expired card and other bank cards for transformation, re business card printing or magnetic writing.

4. Obtain a new card after reporting the loss as a real cardholder.

The risk of credit card fraud is increasing day by day

According to the “China bank card industry development report (2017)” released by China UnionPay, the main fraud means of debit card from 2016 to 2017 are telecommunication fraud, Internet fraud and counterfeit card fraud. In the credit card fraud, the proportion of counterfeit card fraud means decreased, but the false card fraud loss still ranked first.

How to help banks build anti fraud model

In addition to transaction fraud, the risks of application fraud and merchant fraud can not be ignored. The former is that illegal elements illegally apply for bank card through false identity or information; in merchant fraud, illegal cash out through the transaction of credit card on POS machine, using card to wash bills, collusion to accept fake card and other behaviors are also common. They often take advantage of the loopholes of credit card transactions and seek profits by charging a certain share of commission.

These behaviors will not only bring direct economic losses, but also bring great risks to the normal business operation of banks.

Top image helps banks build credit anti fraud model

In the cross-border e-commerce payment transaction with credit card collection as the payment method, the risk of fraud transaction is prominent. In addition, counterfeit card fraud, false application, Internet fraud and other acts are also important fraud risks faced by the bank.

The bank uses the top image intelligent analysis platform to build a risk control model based on business requirements, which effectively improves the anti fraud ability of transactions.

First of all, the intelligent analysis platform is used to clean the bank’s million level data. By deleting meaningless fields, the data with high missing rate, single value or identical distribution are cleaned. The scale of data is reduced, which lays the foundation for rapid data analysis and modeling.

Then, preprocess the data, extract the available features from the massive data, and establish the large width table. In the construction of the wide table, the merchant, the individual, the merchant and the individual are taken as the summary objects, and the transaction times, transaction amount, transaction currency, transaction time interval, number of different merchants and total transaction times are taken as fields, and different time segments are selected as time windows to realize the summary of various statistical data.

Through data preprocessing, the following salient features are sorted out

1. The overall fraud rate in the transaction is 2% – 3%.

2. Most transaction frauds are initiated by non cardholders themselves, and a large number of cards are used in a short period of time, involving a large amount of money.

3. The fraud rate of merchants with 120-150 transactions reached 50%. With the increase of the number of transactions, especially for the merchants with more than 2000 transactions, the possibility of merchant fraud will be reduced. Merchants with less than 150 transactions should focus on monitoring.

4. In the given data, there are still very few merchants with thousands of transactions and 1000 + fraudulent transactions. This shows that the current fraud detection methods are not mature and can not find suspicious merchants in time.

5. From the perspective of fraud rate, merchants with fewer transactions have higher fraud rate and are more prone to fraud.

Based on the above data, the top image intelligent analysis platform derived the important features of several thousand dimensions. According to a certain proportion of training and verification data sets, the appropriate machine learning model was used to make the AUC (area under the curve) of the model above 0.75. Then, the data of model training is adjusted, and the samples are sampled and adjusted to enter the model training. When the accuracy rate reaches more than 90%, 70% of fraudulent transactions can be covered.

After the model training, the data in the front of the prediction list is selected as the final prediction list, which achieves a high hit rate.

Top image intelligent analysis platform covers data transmission, data storage, data management, data ETL, AI modeling and other functions, which can greatly reduce the threshold for enterprises to use artificial intelligence technology. Enterprises only need to provide relevant data to realize data ETL, modeling and other operations. Through offline scheduling, real-time analysis and decision-making, data analysis and application are accelerated, and enterprise application is accelerated New AI technology.

Recommended reading

Application of new generation flow computing framework in financial technology
WeChat, electricity supplier, air ticket, and tiktok on the trill are making “business”.