NextFin News - Yandex has deployed a new neural network architecture capable of identifying fraudulent phone calls in real-time, marking suspicious incoming traffic before a user even answers the device. The system, unveiled by Alexander Lunev, head of information security training at Yandex, represents a significant shift in the arms race between cybersecurity firms and organized fraud syndicates using "SIM boxes" to automate mass-dialing campaigns.
The technical backbone of the feature relies on the simultaneous analysis of over 300 behavioral factors. Unlike traditional caller ID services that depend on static blacklists or delayed user reports, the Yandex algorithm evaluates network-level anomalies such as call frequency, duration, and the specific patterns of mass-dialing characteristic of GSM gateways. According to Lunev, the system can process these variables and render a "suspicious" verdict in less than 60 seconds, effectively tagging the call on the smartphone screen as the phone is still ringing.
The urgency of this deployment is underscored by deteriorating security metrics. Data from the Ministry of Internal Affairs indicates that telephone fraud cases surged by 23% in 2025, with total financial damages exceeding 30 billion rubles. The rise of SIM boxes—hardware that allows scammers to cycle through hundreds of spoofed numbers in minutes—has rendered manual blocking nearly obsolete. By identifying the infrastructure behind the call rather than just the number itself, Yandex aims to neutralize the advantage of number spoofing.
However, the reliance on behavioral AI introduces a layer of technical friction. While the system uses anonymized data to preserve privacy, the "black box" nature of neural network decision-making carries the inherent risk of false positives. A legitimate high-volume caller, such as a delivery service or a medical clinic, could theoretically trigger the same behavioral flags as a fraudulent entity if their calling patterns mimic the rapid-fire cadence of a SIM box. Lunev emphasized that the AI serves as a "routine filter" rather than an absolute gatekeeper, noting that the final decision to engage remains with the human user.
From a market perspective, Yandex’s move is a defensive play to retain its ecosystem's utility as mobile network operators (MNOs) face increasing pressure to implement similar network-level protections. While the AI provides a robust barrier against mass-market scams, it remains less effective against highly targeted "social engineering" attacks where the caller’s behavior may appear perfectly normal to a statistical model. The effectiveness of this tool will ultimately depend on its ability to evolve faster than the scammers, who are already experimenting with AI-generated voice clones to bypass traditional detection methods.
Explore more exclusive insights at nextfin.ai.
