Anomaly detection in computer networks using type-2 fuzzy logic.

摘要

This dissertation presents a novel approach in computer networks that applies a new type of Fuzzy Logic, referred to as Type-2, to the detection of intrusions in computer networks. Type-2 Fuzzy Logic was developed to handle uncertainties inherent in many applications of traditional Fuzzy Logic, which is generally referred to as Type-1.; Fuzzy Logic concepts have previously been applied to Intrusion Detection, for both anomaly and misuse detection. Type-1 Fuzzy Logic, when compared with other solutions, produces detection rates and false alarm rates that are comparable to existing intrusion detection systems. However, Type-2 Fuzzy Logic, when applied to anomaly detection as a complement to Type-1, shows significant improvements in detection rates for network-based intrusions. Improvements are accomplished by reducing the uncertainties that appear in the rules used to evaluate the Type-2 data sets that represent network data, and in the creation of Type-2 membership functions.