Research On Computational Intelligence In Intrusion Detection

With the rapid development of Internet technologies, the applications of networkare also becoming more widely, As a result, network protection has become a specialurgent and important thing. Traditional intrusion prevention technologies, such asfirewalls, data encryption and virtual private network are all passive preventiontechnologies, they have failed to fully protect networks and systems when facing withincreasingly sophisticated attacks and malicious software. As a kind of positive defensetechnology, Intrusion detection systems (IDS) can detect threats before they causedamage to the system and make a response, which effectively made up for the defect ofthe traditional defense technologies. Therefore, intrusion detection systems (IDS) havebecome an indispensable part of security products.The paper mainly studies the particle swarm optimization algorithm(computational intelligence technology) and self-organizing maps (SOM) neuralnetworks of computational intelligence and the applications in intrusion detection, thedetails are as follows:①By introducing the chaos mechanism into particle swarm optimization (PSO)algorithm, the convergence of the particle swarm optimization (PSO) algorithm isinvestigated. We propose a new chaos particle swarm optimization algorithm (CPSO),in which we use a premature judgment mechanism to determine whether the PSOalgorithm is getting into a local optimum, if so, chaos researching can start a newsearching. By doing so, the algorithm can solve the premature problem. Finally, weuse the CPSO algorithm into intrusion detection.②By studying the diversity of the particle swarm, a new multi-swarm particleswarm optimization algorithm (MPSO) based on mixed search behavior is proposed. Bymeans of sub-populations with mixed search behavior, MPSO can remain the diversityof the swarm, based on which an adaptive re-initialization mechanism is introduced inthe algorithm. When the diversity of the population is below a predefined threshold, wecan believe the algorithm has got into a local optimum. Then we need to re-initialize theposition of particles to start a new search in the new area, through which the algorithmcan convergence to the global optimum. MPSO is then successfully applied to intrusiondetection. ③We focus on the self-organizing maps (SOM) neural network, by using theCPSO algorithm to choose the initial connection weights of the SOM, the algorithm hasovercome the sensitivity to initial values. Thus, the SOM with the CPSO algorithm(CPSO-SOM) can be used in intrusion detection.