DEVELOPMENT AND USE OF NEURAL NETWORKS FOR PURPOSE OF SECURE INFORMATION PROCESSING

Abstract

The using encryption, private information is protected from prying eyes. These increasingly complicated and computationally intensive encryption algorithms exist because there are so many of them. For security, neural networks and encryption can be utilized together. These systems can be improved by using neural network architectures to detect anomalies, which can then distinguish between benign and malicious packets. A normalized and selected/reduced feature set is typically pre-processed onto the data before it is fed into the IDS in order to effectively predict anomalies. Training a neural network to optimize its performance takes into account the effects of pre-processing techniques. Described in this paper are the algorithms involved in pre-processing.