Identifying data streams anomalies by evolving spiking restricted Boltzmann machines

Abstract

Data streams are characterized by high volatility, and they drastically change in an unpredictable way over time. In the typical case, newer data are the most important, as the concept of aging is based on their timing. These flows require real-time processing in order to extract meaningful information that will allow for essential and targeted responses to changing circumstances. Knowledge mining is a real-time process performed on a subset of the data streams, which contains a small but recent part of the observations. Timely security requirements call for further quest of optimal approaches, capable of improving the reliability and the accuracy of the employed classifiers. This research introduces a real-time evolving spiking restricted Boltzmann machine approach, for efficient anomaly detection in data streams. Testing has proved that the proposed algorithm maximizes the classification accuracy and at the same time minimizes the computational resources requirements. A comparative analysis has shown that it outperforms other data flow analysis algorithms.