Network Intrusion Detection Using an Evolutionary Fuzzy Rule-Based System

The proliferation of computer networks has brought network security to the forefront. It has become imperative to devise new methods for network intrusion detection. Current methods are frequently unable to detect polymorphic or novel attack modes. In addition, the enormous volume of network traffic makes it difficult to monitor and evaluate all features of communication packets on the network. This research uses evolutionary learning to select key traffic features to be used in intrusion detection. An evolutionary algorithm is then used to learn a set of fuzzy inference rules based on the selected features. The resulting system is able to adapt to dynamic network conditions and indentify morphing and unknown attack modes. Tests show a very high success rate in identifying attacks with minimal false positives. Keywords-evolutionary learning; network security; network intrusion detection; fuzzy logic