DISTRIBUTED SELF-ORGANIZED SYSTEM FOR PREDICTING MALICIOUS ACTIVITY IN COMPUTER NETWORKS

Abstract

In this article, a self-organized computer network protection system based on deep learning algorithms was considered In addition, a new self-organizing incremental neural network called FG-SOINN written in the Python programming language was presented. In the SOINN, node and edge removal is defined by two parameters that need to be optimized for each existing program using cross-validation or similar resampling approaches. FG-SOINN overcomes this drawback by treating node and edge removal as an integral part of the learning process. Three concepts were formulated to form "garbage oblivion": idle time, reliability, and utility by which the network removes nodes and edges. Such a network is based on the concept of "learning without a teacher" and will work both with artificial and real data and even with sudden or repeated deviationst.