Precision and Safety: A Russian Neural Network for Security Screening Learns From Domestic Experience
Diagnostika-M, part of Russia’s National Security Technology System, has completed development and unveiled an updated version of its neural network for TS-SCAN inspection scanners.
A Knowledge Base Built in the Field
These systems, designed to screen baggage and cargo, are a core element of transport security infrastructure. They are deployed at railway stations, airports, border crossings, and other high-traffic public locations. Each year, inspection scanners are equipped with increasingly advanced “intelligent” components, a trend that is set to reshape security screening standards across Russia.
In 2024, Diagnostika-M, a resident of the Technopolis Moscow special economic zone, launched serial production of inspection equipment powered by artificial intelligence. At the initial stage, the neural network was trained on 5.5 million images of dangerous items. That dataset was later expanded to 8.5 million images. As a result, operators can now reliably distinguish real weapons from replicas, toys, or everyday objects with similar silhouettes.
The neural network is capable of identifying an expanded range of threats – from standard firearms to ammunition, explosives, and individual explosive components. Crucially, it performs this recognition across multiple viewing angles and under complex conditions where objects overlap inside baggage. The underlying dataset is not a laboratory construct but a vast collection assembled through real-world operation at Russian transport facilities. This practical foundation is the system’s key competitive advantage.
A Special Watch Zone
Improved screening accuracy directly translates into a higher level of public safety. The risk of false alarms that slow down procedures is reduced, and, more importantly, so is the probability of missing genuinely dangerous items. The technology has completed its development phase and is ready for industrial trials. In the near term, serial production of TS-SCAN scanners with the upgraded AI is planned, marking a tangible step toward the digital transformation of security infrastructure.
From the outset, the Russian development has been tailored to domestic market requirements. This focus is both a strength and a limitation for export. Direct overseas sales are unlikely, as each market requires models trained on its own datasets and threat profiles. However, the methodology, platform architecture, and expertise developed by Diagnostika-M may be in demand among countries willing to invest in building localized security systems of their own.
A Proven Version
Over the next few years, intelligent inspection scanners of this type are expected to be rolled out widely across Russian facilities. Further development may follow a modular path, with neural network solutions adaptable to the specific needs of different environments – from metro systems to international airports. In the longer term, the system could become the core of more complex digital platforms that aggregate data from multiple sensors to deliver real-time, integrated threat analysis.
The presented neural network demonstrates the maturity of Russia’s security-focused development ecosystem, which prioritizes deep technological adaptation for concrete, mission-critical national tasks.
