CAJ | 학술논문

SVM有着很强的学习能力，已经成为入侵检测的重要算法之一。由于入侵检测原始数据量大，且具有高维性、冗余性等特点，导致传统SVM入侵检测算法计算量大、预测时间长。基于此，文中提出一种改进的SVM入侵检测算法( KP-CA-GA-LC-SVM)。文中利用核主成分分析法( KPCA)进行数据的特征提取，降低数据维数和计算量；使用两个核函数线性加权结合形成的组合核函数代替传统的单一核函数，并通过遗传算法( GA)进行SVM核参数及组合核权系数的寻优，来提高SVM性能。实验结果表明，文中算法有效地提高了入侵检测的检测精度。
SVM유착흔강적학습능력，이경성위입침검측적중요산법지일。유우입침검측원시수거량대，차구유고유성、용여성등특점，도치전통SVM입침검측산법계산량대、예측시간장。기우차，문중제출일충개진적SVM입침검측산법( KP-CA-GA-LC-SVM)。문중이용핵주성분분석법( KPCA)진행수거적특정제취，강저수거유수화계산량；사용량개핵함수선성가권결합형성적조합핵함수대체전통적단일핵함수，병통과유전산법( GA)진행SVM핵삼수급조합핵권계수적심우，래제고SVM성능。실험결과표명，문중산법유효지제고료입침검측적검측정도。
SVM has a strong learning ability,has become one of the most important intrusion detection algorithm. Due to a large amount of raw data in intrusion detection,and with a high dimension,redundancy,etc. ,result in larger of calculating the volume and the longer of predicted time in the traditional SVM intrusion detection algorithm. Based on this,propose an improved SVM intrusion detection algo-rithm ( KPCA-GA-LC-SVM) . In this paper,use Kernel Principal Component Analysis ( KPCA) for data feature extraction and reduce the dimensionality of data and computation. Use a combination of kernel functions formed by weighted linear combination of two kernel function instead of the traditional single kernel function,and through genetic algorithm to find the optimization of kernel parameters and the weights of the composed kernel function to improve the performance of SVM. The experimental results show that the improved algo-rithm can effectively improve the accuracy of intrusion detection.