Font Size: a A A

Research On Methods Of Intrusion Detection System Based On Fuzzy Clustering Algorithms

Posted on:2007-01-31Degree:MasterType:Thesis
Country:ChinaCandidate:H LuFull Text:PDF
GTID:2178360215975963Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Intrusion detection system which succeeds to the traditional safe measures, such as firewall, data encryption and so on is a kind of new safe protective technology. As a kind of active safe technology, Intrusion detection system has recently been researching hot. At present, the research of the methods of intrusion detection is mainly unfolded around the misuse detection and the anomaly detection. The misuse detection exists in the problem of inefficiency and being unable to detect unknown attacks and variant attacks etc. The anomaly detection exists in the problem of the inability of building model and low detection rate etc. This paper which is based on deeply researching the present methods of intrusion detection and clustering analysis algorithms has realized a prototype of fuzzy clustering intrusion detection system through presenting two new fuzzy clustering algorithms and introducing the method of fuzzy clustering analysis in the method of the anomaly detection of intrusion detection.The major achievements in this paper can be generalized into two implementations and two new algorithms:1. A software which can capture network data parcels and extract feature from them has been implementedThe software can extract useful features from network flow data captured by retreating and conserve them in the form of record set by recombining and arranging and provide the upper intrusion detection with them. Not only can this software be used as a pretreatment software, but also it can be used alone as a network monitoring software.2. A system of fuzzy clustering analysis and anomaly detection has been implementedThis system has integrated several typical clustering algorithms and two new fuzzy clustering algorithms presented in this paper, and implemented the pretreatment of data congregation and data based on primary component analysis. Moreover, this system is a process of integral anomaly examination that can implement the training of intruding data congregation,and that can form normal or abnormal mode library and that can detect the record of unknown data congregation.3. A new possibility fuzzy clustering algorithm based on the uncertainty membership has been implementedThis paper has presented a theory of the uncertainty membership and a new algorithm based on the theory, namely, possibility fuzzy clustering algorithm based on the uncertainty membership. This paper has given the comparable whole description and analysis. The results of the experiments based on the datum of KDDCUP99 demonstrate that the algorithm possesses the upper detection rate and the lower misuse detection rate.4. A new genetic tabu fuzzy clustering algorithm has been implemented Traditional fuzzy clustering algorithm based on objective function is an iterative hill-climbing algorithm and easy to fall into local optimization. This paper put forward the fuzzy clustering algorithm based on genetic tabu algorithm, which applies synthetically many springboards of genetic algorithm and memory property of tabu algorithm to improve the clustering effect and produces the optimal clustering center by using iterating genetic tabu search algorithm.
Keywords/Search Tags:Intrusion Detection, Fuzzy Clustering, Possibility Membership Degree, Uncertainty Membership Degree, Genetic Algorithm, Tabu Search
PDF Full Text Request
Related items