METHOD OF INTELLIGENT SIMULATION OF TOPOLOGICAL CHARACTERISTICS OF COMPUTER NETWORKS IN NETWORK RECONNAISSANCE

The purpose of the study: development of a method that includes classes of random graph models, metaheuristic algorithms for numerical optimization, and generative machine learning models, designed to solve the problem of intelligent imitation of the topological characteristics of computing networks in order to counteract cyber reconnaissance by intruders, taking into account the algorithmic complexity and resource limitations of security measures. Methods used: weighted ideal point method, random graph model, modified genetic algorithm of numerical optimization, Bayesian optimization algorithm, convolutional variational autoencoder model, backpropagation method. The result of the study: the presented method allows us to ensure counteraction of computer intelligence by demonstrating to the attacker false topological characteristics of the computing network by optimizing the number and connectivity indicators of false and real nodes, taking into account the requirements for resource constraints on the creation and maintenance of false information directions and nodes. Scientific novelty: consists in the development of a method for intelligent simulation of the topological characteristics of computational networks, which differs from the known method by using a modified genetic algorithm and a random graph model based on binary random vectors of variable length to form a training array, and by applying a Bayesian algorithm to select the optimal point for synthesizing a false topology of a computational network from the hidden space of a trained variational autoencoder, which corresponds to the minimum Cartesian distance from the ideal point to the current point in the weighted criterion space.