METHOD AND SYSTEM FOR PREVENTING MALICIOUS ATTACKS ON NETWORK INFRASTRUCTURE USING CASCADES OF SELF-TEACHING NEURAL NETWORKS Russian patent published in 2025 - IPC G06F21/55 

FIELD: information security.

SUBSTANCE: in the method of preventing malicious attacks on network infrastructure using cascades of self-teaching neural networks, using a computing device of the system, forming a cascade of neural networks connected by layers; forming the self-training feedback for the neural network stages layers relative to the obtained statistical data and the packet information repeating sequences; groups of statistical data are supplied to the first layer of the cascade of neural networks; generating data for abnormal zones at second layer of neural network cascade; grouping statistical data in a multidimensional field on the third layer of the neural network cascade; evaluating the output data and data obtained as a result of the operation of the previous layer, and classifying the group for belonging to the type of attack on network infrastructure on the fourth layer of the neural network cascade.

EFFECT: high accuracy and speed of preventing malicious attacks on network infrastructure, including zero-day attacks – previously unknown targeted attacks and specially prepared group malicious intrusions.

13 cl, 14 dwg, 6 tbl