METHOD OF IDENTIFYING AND BLOCKING MALICIOUS SOFTWARE USING CASCADES OF SELF-LEARNING NEURAL NETWORKS Russian patent published in 2025 - IPC G06F21/56 G06N3/08 

FIELD: computer engineering.

SUBSTANCE: result is achieved due to a method in which: obtaining information on a software launch instruction and controlling the machine code execution process; forming groups for which statistical information on software operation is extracted and calculated; identifying additional data on behaviour of controlled software; detecting patterns; forming a cascade of neural networks; generating self-training feedback for layers of cascades of neural networks; based on the calculated parameters, training the neural network in order to identify the malware; using a trained classifier to determine whether the software belongs to a certain family of programs; sending groups of statistical data to form node data structures; generating data for abnormal zones; grouping statistical data in a multidimensional field; forming a number of expert classifiers in the neural network cascade; classifying the group as to belonging to the type of malware; blocking malware.

EFFECT: high accuracy and speed of identifying and blocking malicious software.

13 cl, 2 tbl, 3 dwg