Naive Bayes Classifiers using Principal Component Analysis for Intrusion Detection - Based on p-value

• Main Authors: Wei tung Tsai, 蔡煒彤
• Other Authors: Wei-Ning Yang
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/70353694698999524317

碩士 === 國立臺灣科技大學 === 資訊管理系 === 102 === Naive Bayes classifier is a simple probabilistic classifier which applies Bayes' theorem based on strong (naive) independence assumptions between the features to avoid the curse of dimensionality. We first apply principal component analysis to obtain the uncorrelated transformed features and then apply Naive Bayes algorithm based on the transformed features. The p-value associated with each transformed feature of the testing instance is evaluated based on the distribution of the corresponding transformed feature estimated from the training dataset. Based on Naive Bayes independence assumptions, the joint p-value for each testing instance is evaluated for Bayesian classification. The proposed hybrid algorithm is evaluated through the accuracy for detecting anomaly-based intrusion on NSL-KDD dataset. The experimental results demonstrate that principal component analysis can (substantially) increase the detection accuracy of the Naive Bayes classifier.