METHOD OF DETERMINING ACTUAL THREATS TO INFORMATION SECURITY OF INFORMATION INFRASTRUCTURE OBJECTS BASED ON WORD VECTORIZATION METHOD IN ARTIFICIAL NEURAL NETWORK Russian patent published in 2025 - IPC G06F21/57 G06N3/02 

FIELD: information security.

SUBSTANCE: method of determining actual threats to information security of information infrastructure objects based on a method for vectoring words in an artificial neural network comprises steps of obtaining a list of known vulnerabilities and threats, attack patterns; creating an artificial neural network (ANN), which takes values of vulnerability features, with hidden layers of ANN, including layers of vector representations of words, long-short-term memory and fully connected layers, with one output, which forms a list of names of actual threats with given vulnerability features; obtaining a set of training samples of vulnerabilities, threats and patterns of computer attacks; training created ANN with verification on reference data by input of training samples; obtaining a list of detected vulnerabilities; vulnerability features are sent to the input of the created ANN and a list of actual threats to the information infrastructure object is obtained at the output.

EFFECT: reduced complexity and labour intensity of the process of forming a list of actual threats to information infrastructure objects.

1 cl, 2 dwg