Banking fraud detection is not possible without in-depth data analysis in real time


In mid-September, the new European Payment Directive came into force, requiring Strong Customer Authentication. Electronic payment service providers have had to introduce new authentication elements for transaction authorization by clients.

In mid-September, the new European Payment Directive came into force, requiring Strong Customer Authentication. Electronic payment service providers have had to introduce new authentication elements for transaction authorization by clients. This measure has undoubtedly increased the security of operations, but it will not prevent fraudsters from conducting further cyber-attacks against bank accounts. Banks must continue strengthening their defenses, invest in their ICT systems and leverage a wider range of data sources for risk management.



To be able to process the growing volumes of data exchanged with the customer, the merchant, their bank, and the payment gateway, banks need high-performance computing resources with advanced software to process all relevant data in full context and detect potential threats. This requires the deployment of real-time technologies with deep analytics that process vast amounts of data and perform very complex operations in real time. They use all available data that the customer has granted consent for, such as transaction history, centrally managed watchlists, and other external databases and statistics, which they combine. Deep analysis can quickly test multiple methods to determine which models work best. Such systems are able to link millions of pieces of information to detect anomalies and respond appropriately within seconds. Neural network models use machine learning without human intervention. The algorithms continuously modify themselves with updated data, use previous results for new analyses, and improve with every interaction. They monitor how the user interacts with the system to detect anomalies that could indicate identity theft or malware. Rel-time fraud detection generates fewer false positives without compromising the quality of the customer experience. For instance, when verifying that a card payment is being made by the actual cardholder, the system compares the payment details with the locations, times, amounts, and subjects of previous payments. It verifies the device where the payment originates and checks the beneficiary’s account. And if the system detects any irregularities or anomalies, it will terminate the operation and forward it for review.



Fraudsters have been improving their capabilities to penetrate and control or circumvent risk management rules defined in the system. It is therefore essential to continuously perform adaptive behavioral analytics to capture and assess behavioral patterns in each transaction and detect new risks and unknown fraud patterns. Creates a deep history profile for the bank account, cardholder, customer, merchant, PoS terminals, web session, etc. The more available profiles the system evaluates, the better it can understand whether a payment or other type of transaction is legitimate. It can identify potentially fraudulent behavior before a payment is made or a customer’s account is compromised.



To make the fraud detection system more effective, you need to properly define the workflow and its priorities. You need to sort data by risk level, create groups and standards, and aggregate tasks by subject. This reduces the volume of queries to the source systems and the number of widgets to be processed. It also gives analysts a quick and more comprehensive view of subject behavior and automatically assigns alerts to investigators based on pre-defined rules. 



Enforcing stronger authentication is just one small element in countering cybercrime. To effectively combat this, you need to monitor incident reports, social media, cyber events, and other available data to help uncover previously unknown risks. Leverage contextual information hidden in text, image, and audio content and connect clues from a multitude of seemingly disparate data. We can expect convergence of electronic data with behavioral profiles to continue. Real-time prevention will be faster and more scalable and will enable higher transaction throughput. More extensive sharing of information between banks, other corporations and institutions that will newly implement fraud prevention and detection systems will also be important. This will allow to correlate billions of daily transactions and enrich the data with business context. This will create sophisticatedly structured data that can be analyzed in many different combinations across products and organizations. The result will be a perfect overview of current security risks in real time, with event prediction and suggestions for optimal solutions.

Radek Basár, ICT development manager, Komerční banka

Necháte nám drobečky, abychom věděli, jak nám to tu žije?

Ale jo

Vyberte si sušenky přesně podle vaší chuti a nálady.

Funkční cookies jsou nezbytné, aby nám web fungoval tak, jak má. Bude si tak pamatovat i nastavení a souhlasy, které jste nám udělili.

Souhlasím se všemi