Sparse Least Squares Support Vector Clustering Algorithm for Anomaly Detection
Hao Xiong, Sheng Sun · International Journal of Applied Mathematics & Statistics/International journal of applied mathematics and statistics · 2013
Intrusion detection system has become an important part of network security technology. The major benefit of anomaly detection algorithms is their ability to potentially detect unforeseen attacks. In this paper, we present a Least Squares Support Vector Clustering (LSSVC) algorithm first. LSSVC has the advantages which can cluster mess of data and identify noise. But this algorithm has the problem that sparseness is lost. In view of its shortcomings, then we utilize a pruning method to clip support vector set, and propose a new sparse LSSVC. Recognition precision and recognition speed are improved after pruning. The experimental results show that our new algorithm has high feasibility and validity, and it will have vast potential for future development in the field of intrusion detection.