计算机工程
計算機工程
계산궤공정
COMPUTER ENGINEERING
2013年
11期
131-135
,共5页
入侵检测%混沌粒子群优化算法%最小二乘支持向量机%联合优化%特征选择%混沌机制
入侵檢測%混沌粒子群優化算法%最小二乘支持嚮量機%聯閤優化%特徵選擇%混沌機製
입침검측%혼돈입자군우화산법%최소이승지지향량궤%연합우화%특정선택%혼돈궤제
intrusion detection%Chaotic Particle Swarm Optimization(CPSO) algorithm%Least Square Support Vector Machine (LSSVM)%joint optimization%feature selection%chaotic mechanism
为提高网络入侵检测效果，提出一种结合混沌粒子群优化(CPSO)算法和最小二乘支持向量机(LSSVM)的网络入侵检测模型。将网络特征和 LSSVM 参数编码成二进制粒子，根据网络入侵检测正确率和特征子集维数权值构造粒子群目标函数。通过粒子群找到最优特征子集和 LSSVM 参数，同时引入混沌机制保证粒子群的多样性，防止早熟现象的出现，从而建立最优网络入侵检测模型。采用KDD99数据集进行性能测试，结果表明，该模型不仅能获得最优特征子集和LSSVM参数，而且提高了入侵检测速度和正确率，降低了入侵检测误报率和漏报率。
為提高網絡入侵檢測效果，提齣一種結閤混沌粒子群優化(CPSO)算法和最小二乘支持嚮量機(LSSVM)的網絡入侵檢測模型。將網絡特徵和 LSSVM 參數編碼成二進製粒子，根據網絡入侵檢測正確率和特徵子集維數權值構造粒子群目標函數。通過粒子群找到最優特徵子集和 LSSVM 參數，同時引入混沌機製保證粒子群的多樣性，防止早熟現象的齣現，從而建立最優網絡入侵檢測模型。採用KDD99數據集進行性能測試，結果錶明，該模型不僅能穫得最優特徵子集和LSSVM參數，而且提高瞭入侵檢測速度和正確率，降低瞭入侵檢測誤報率和漏報率。
위제고망락입침검측효과，제출일충결합혼돈입자군우화(CPSO)산법화최소이승지지향량궤(LSSVM)적망락입침검측모형。장망락특정화 LSSVM 삼수편마성이진제입자，근거망락입침검측정학솔화특정자집유수권치구조입자군목표함수。통과입자군조도최우특정자집화 LSSVM 삼수，동시인입혼돈궤제보증입자군적다양성，방지조숙현상적출현，종이건립최우망락입침검측모형。채용KDD99수거집진행성능측시，결과표명，해모형불부능획득최우특정자집화LSSVM삼수，이차제고료입침검측속도화정학솔，강저료입침검측오보솔화루보솔。
In order to improve the network intrusion detection effect, this paper puts forward a network intrusion detection model based on Chaotic Particle Swarm Optimization(CPSO) algorithm and Least Squares Support Vector Machine(LSSVM). The network features and parameters of LSSVM are encoded into binary particles. The objective function of particle swarm optimization algorithm is built based on network intrusion detection accuracy and the dimensions of the feature subset. The particle swarm is used to find the optimal feature subset and LSSVM parameters, while the chaotic mechanism is introduced to guarantee the diversity of particle swarm and to prevent producing precocious phenomenon, the optimal model of the network intrusion detection is established. The performance of proposed model is test by KDD99 data and the simulation results show that proposed model can select the optimal feature subset and LSSVM parameters, the detecting speed and network intrusion detection accuracy are improved, and thereby network intrusion detection false negative rate and false positive rate are reduced.
