A neural network developed by Russian researchers can automatically identify suspicious financial transactions to help prevent fraud, according to TASS.
The project is associated with Peter the Great St. Petersburg Polytechnic University (SPbPU). The system relies on deep learning and considers a range of factors, including the card type and card number, information about the cardholder, data about the recipient, and how the transfer was executed, whether through a banking app or another channel. It also analyzes historical transaction patterns to uncover subtle shifts that may indicate fraud, offering a more nuanced view of money movement.
As one scenario explained by the researchers, if a person opens a bank account and maintains a daily average of about 1,000 rubles for six months, then suddenly receives a transfer of 30,000 rubles in a single day, the model’s probability score for fraud would rise, prompting closer monitoring and further verification.
During testing, the system demonstrated strong performance and showed readiness for practical deployment. The creators contend that the neural network could serve as an initial safeguard for bank customers, helping to detect unlawful actions and reduce risk across financial networks.
In related news, some ancient studies described a distinct type of activity in the brain of a dying person, a finding unrelated to the financial fraud detection work described above and noted here for historical context.