An expert on neural networks spoke about the methods of AI's fight against fraudsters

In the digital age, where money has become a data stream, its security has become the number one issue. However, in modern banking systems, artificial intelligence (AI) algorithms create an accurate digital profile of each customer, recording their typical financial habits, from the average check to the time and place of transactions. Dmitry Titov, head of the Quality Management and Artificial Intelligence Implementation Department at AKFIX-RUS, told Izvestia on December 12.

"In case of any deviation from the established pattern, for example, when attempting a large transfer at a non-standard time or region, the system instantly blocks the operation for additional verification and sends a notification to the user. This intelligent analysis of hundreds of parameters in real time allows us to stay ahead of the actions of scammers," he explained.

Special attention is paid to the fight against social engineering. Algorithms trained on thousands of hours of recordings of conversations with scammers have learned to recognize characteristic speech patterns: pressure on urgency, specific formulations or SMS code requests. In pilot projects, the systems warn the subscriber about a suspicious call right during the conversation.

The investigation was introduced: "AI detective" will deal with a difficult task with a lack of data

How will the new approach help to rid neural networks of errors that occur when information is scarce?

Access to financial instruments is now protected by a multifactor authentication system, where AI provides accurate biometric identification by face or voice through a single biometric system and monitors unauthorized access attempts around the clock. For online payments, the algorithm evaluates the risk of each transaction and decides whether SMS confirmation is sufficient or additional verification is required.

Protection also works outside of banking ecosystems: when paying on marketplaces, AI checks the reputation of sellers, and browser extensions block traffic to phishing lookalike sites. Even in messengers, systems are being implemented that warn about typical cheating schemes.

"The result of complex systems is round-the-clock protection without active user participation, instant notifications of suspicious activity, the ability to lock a card with one click and reduce the risk of losing funds by 70-80%. In the future, AI will continue to evolve towards creating adaptive security ecosystems that will automatically adapt to user behavior, age, and digital literacy," Titov said.

One of the key vectors of development will be the personalized protection of vulnerable groups, such as the elderly or children, for whom special modes with enhanced warnings or control of in-game purchases can be created. The next step will be the development of proactive voice protection capable of analyzing a fraudster's voice in real time and warning the subscriber before any information is disclosed, which in the future may become a standard built into smartphone operating systems.

On December 8, the Department for combating the illegal use of information and communication technologies (UBK) of the Russian Interior Ministry reported that fraudsters began disguising themselves as employers, offering "dream jobs" and luring citizens out of personal data and financial information. It was clarified that the scheme begins with a tempting offer of remote work without experience and with a high daily income. After the response, communication is immediately transferred to the messenger, which in itself is atypical for official vacancies.

All important news is on the Izvestia channel in the MAX messenger.