Title :
Intrusion Detection Algorithm Based on Semi-supervised Learning
Author :
Li, Yongzhong ; Li, Zhengjie ; Wang, Rushang
Author_Institution :
Sch. of Comput. Sci. & Eng., Jiangsu Univ. of Sci. & Technol., Zhenjiang, China
Abstract :
In order to overcome the shortage that intrusion detection system is sensitive to outlier, we propose an intrusion detection algorithm based on semi-supervised fuzzy clustering. In this algorithm, the training data for semi-supervised learning is a hybrid data of labeled and unlabeled samples. While training the system model, we use a few labeled samples and many unlabeled samples as seeds initializing the classifier of the system. Under the constraint of labeled data, we use fuzzy C-Means method to generate clusters without many labeled data and uneasily plunges locally optima. Comparing with FCM algorithm, the experiment results on data sets KDD CUP 99 has shown the effectiveness of the proposed algorithm, it has higher detection rate and lower false detection rate.
Keywords :
fuzzy set theory; learning (artificial intelligence); pattern clustering; security of data; FCM algorithm; fuzzy c-means method; intrusion detection algorithm; semi-supervised fuzzy clustering; semi-supervised learning; Algorithm design and analysis; Classification algorithms; Clustering algorithms; Intrusion detection; Supervised learning; Testing; Training data; Semi-supervised learning; fuzzy clustering; intrusion detection;
Conference_Titel :
Information Technology, Computer Engineering and Management Sciences (ICM), 2011 International Conference on
Conference_Location :
Nanjing, Jiangsu
Print_ISBN :
978-1-4577-1419-1
DOI :
10.1109/ICM.2011.197