Author(s): Shulha, O., Yanenkova, I., Kuzub, M., Muda, I., & Nazarenko, V.
The following groups of banks` cyber threats have been identified: network and application layer attacks; social engineering; developed stable threats; cybercrime, master data violation. Banks are suffering from phishing attacks and bank roses, which is the most common mobile cyber threat, since most of smart phones holders also have a bank card. A neural network was created, consisting of a 1st hidden layer with two neurons, moreover, a mathematical interpretation of the source layer wa represented, as well as the 1st and 2nd hidden layers of the neuron. The constructed models were tested for quality and adequacy using the following coefficients: MISC misclassification rates, MSE and ASE error rates. Their results were analyzed and it was found that all the models created were describing the input data with the same accuracy, but nevertheless, the neural network model was the best. The probability prediction of cyber-threat signs occurrence during the transaction of mobile and Internet banking users was performed on the basis of the selected neural network model. As a result, it was found that 21.9% of transactions were likely to contain cyber-threats. The application of this model will enable the banking sector employees to detect cyber threats in transactions, thereby preventing mobile and internet banking users from possible losses caused by certain criminal activity.