Verification of unreliable parameters of the malicious information detection model

The object of research is the process of detecting harmful information in the social networks and global network. There has been proposed the approach to verifying the parameters of a mathematical model of a random process of detecting malicious information with the unreliable, inaccurately (contradictory) given initial data. The approach is based on using stochastic equations of state and observation that are based on controlled Markov chains in finite differences. At the same time, verification of key parameters of a mathematical model of this type — elements of a matrix of one-step transition probabilities — is performed by using an extrapolating neural network. This allows to take into account and compensate the inaccuracy of the original data inherent in random processes of searching and detecting malicious information, as well as to increase the accuracy of decision-making on the assessment and categorization of digital network content to detect and counter information of this class.