Нейросетевой фреймворк ForecaState для защиты промышленности от кибератак создали в СПб ФИЦ РАН

New system demonstrates 30% higher accuracy in threat detection

Scientists from the St. Petersburg Federal Research Center of the Russian Academy of Sciences (SPC RAS) have created the ForecaState neural network framework for detecting cyber threats in industrial enterprises.

The development demonstrated high recognition accuracy on different types of infrastructure — the number of errors in threat detection was 30% less than that of existing analogues.

Deep learning algorithms are suitable for processing large amounts of data, typical for IIoT (Industrial Internet of Things) industrial systems.

IIoT combines "smart" technologies, including sensors, controllers, and robots, that collect data and transmit it to the operator for analysis. This allows optimizing production, reducing costs, improving product quality, and predicting possible equipment failures. In Russia, the technology is actively being implemented in the oil and gas, energy, metallurgical, and machine-building industries, but its use is associated with the risks of cyberattacks and data leaks.

The framework is modular and easily adapts to various production tasks. The system was tested on data from water purification networks and electrical transformers. ForecaState can be used in predictive maintenance systems for early detection of machine failures, load balancing, accident prevention, and quality control on production lines.

Read materials on the topic:

Rostec has come up with a way to protect critical enterprises from ransomware programs

To avoid cyber risks: a guide to secure AI development

They took on security in a Stakhanovite way: Russian companies presented an integrated solution for protecting corporate data

Sources
TASS

Now on home