| Summary: | 碩士 === 國防大學理工學院 === 資訊科學碩士班 === 98 === Due to the impact of information technology, the network intrusion events on the Internet are increased significantly. Hackers often use various hacking techniques to steal privacy information and even to compromise an information system. In this paper, we propose a Genetic Programming Hierarchical Classifier (GPHC) framework, which is able to identify the network events (attack or normal) accurately and efficiently. By using GPHC, the network staff can deal with the anomalous events instantly to avoid the occurrence of damage.
GPHC integrates the Heterogeneous Hierarchical Classifier (HHC) framework and Genetic Programming (GP) theory to obtain better classification results, which is achieved by taking advantages of heterogeneous classifiers and evolutionary computations. Moreover, two algorithms are proposed to enhance the performance of GPHC, namely, GPHC Pruning (GPHCP) algorithm and Strong Base Classifier Generator (SBCG) algorithm. The GPHCP algorithm is used for simplifying classification rules and improving classification efficiency for GPHC. The SBCG algorithm is able to generate high-accuracy classifiers while constructing GPHC. These strong base classifiers can be re-joined to the GPHC constructing phase to improve the classification performance.
GPHC can provide real-time and accurate classification for different problems if the base classifiers are properly chosen. In order to validate the applications of the proposed framework, GPHC has been applied to solve a critical, real-world problem, rainfall intensity classification. Experimental results show that the proposed model is able to achieve high accuracy results which outperform previously published methods.
|
|---|
