Dynamic Evolution Systems and Applications in Intrusion Detection Systems

In this paper, we present a dynamic evolution system and build up a model to trace the transition of the system state. This new model differs from the previous methods, such as Bayesian network, artificial neural network, in two aspects: it can adapt the changes of the environment automatically, and it does not need a special training phase to build up a model. Theoretical analysis shows that it is applicable and practical, and furthermore, experimental results show that it has good performance especially in dynamic environment.