朴素贝叶斯与全贝叶斯在DFI的应用比较研究

A Comparative Study of Naive Bayesian and Full Bayesian in DFI Application

P2P应用软件检测技术主要是深度包检测法和深度流量检测法.深度包检测法不能检测出加密的和未知的对等网络流应用,深度流量检测方法可以克服深度包检测法这个缺点.将模糊识别领域里比较成熟的贝叶斯分类技术应用到对等网络深度流检测中.结合实际项目,对贝叶斯的两个分类器-朴素贝叶斯和全贝叶斯的算法、训练结果、运行结果进行研究,实验研究表明朴素贝叶斯和全贝叶斯分类器能够快速准确地找到P2P流应用,朴素贝叶斯分类器准确度占据优势,全贝叶斯运行时间占据优势.

Deep Packet Inspection and Deep Flow Inspection are the methods that inspect P2P application, Deep Packet Inspection can＇t detect unknown and encrypt P2P application, but Deep Flow Inspection overcomes this shortage. The Bayesian classification technique which is relatively more mature in field of fuzzy recognition is applied to the detection technique of P2P deep flow inspection. Based on practical project,we research Bayesian classifier-Naive Bayesian and full Bayesian in arithmetic, result of training and result of running. The results of an experiment show Naive Bayesian and full Bayesian can rapidly and accurately locate P2P flow application. Naive Bayesian is more advanced in accuracy and full Bayesian is better in runtime.