System for detecting of potentially dangerous communications of network users

Abstract

This work is devoted to the processes of organizing internal information security at the enterprise. The scheme of a software tool for monitoring employee communication and detecting malicious messages using artificial neural network analysis and full-text dictionary search is proposed. A software package designed according to the described scheme has been developed. The scheme of interaction of the program components is considered: a keylogger, a keyboard input analyzer, a user interface, a server coordinating interaction. The schemes of interaction with the analyzer by means of the WebSocket protocol were shown. The interaction of the neural network and the dictionary helped to increase the percentage of accuracy of detecting a dangerous or suspicious message, for example, when the neural network does not consider the text dangerous, and the dictionary considers the opposite, then such text is considered dangerous and is shown to the administrator. Thus, the combination of neural network analysis and dictionary analysis makes it possible to detect more suspicious messages. The work of the software product and its advantages in comparison with other similar systems are demonstrated.