Research And Optimization The Key Technology Of Mail Security Gateway

Email technology is a double-edged sword.The popularity of e-mail has brought great convenience to people's communication.However,the spread of spam also affects and even threatens people's daily life and work.In today's high speed network environment,most email-security gateways are filtering spam through analyzing the content of mail,which is inefficient.The purpose of this thesis is to improve the performance of e-mail security gateways in high speed network environments.The email address matching algorithm and the spam recognition model that do not need to analyze fully the email's are studied in this thesis.First,for email address matching,which involves multi-pattern matching algorithm.Wu-Manber algorithm is a classical multi-pattern matching algorithm,however,the classical algorithm has many limitations in large-scale email-addresses matching.Owing to the failure to consider the email addresses feature,the algorithm is low performance.In this thesis,based on the classical Wu-Manber algorithm and email-address feature,the ME-WM is proposed for large-scale email-addresses matching.The algorithm proposed in this thesis applies a variety of optimization methods to speed up matching performance,has the following major points : choose the best hash algorithm,deal with the hash Collisions by Red Black Tree,match and store the domain of email-address by using Bloom filter.These methods proposed in this thesis reduce the consumption of storage and time of the pattern-matching algorithm.Through experiments,we compare the space consumption and time consumption of ME-WM algorithm and the classical Wu-Manber algorithm and related improved algorithm,for example,RFP-WM and Prefix Tree-WM.And proves that ME-WM algorithm is very suitable for large-scale e-mail address matching.Second,this thesis is the research for spam recognition technology.The Chinese spam recognition model based on the increment of time window is proposed.This model analyses of header spam using decision tree algorithm,explores the association rules between title character and spam,applies of these rules to identify spam.This method has the following advantages: Analyzing the email's title not fully the emails' content,the computation complexity is reduced by this model.The filtering performance of the mail gateway is enhanced;The Chinese segmentation algorithm of open Internet environment is introduced;Constructing the weighted time window is used to calculate each letter to the filtering concept drift score,the concept drift as a key attribute for the construction of the decision tree,There is no need to reconstruct the classifier to solve the concept drift problem;The incremental learning mechanism is designed to expand the spam keyword list for improving the ability to adapt to the dynamic environment model.Finally,the validity of the model is proved by the experiment,and it is superior to the improved Bayesian model,the optimized SVM model and the variable precision rough set decision tree model.