A spam feature selection method and its detection method

Abstract

The invention relates to a feature selection method and a detection method of junk mails. The feature selection method of the junk mails includes the steps that features of the mails are extracted on the basis of a byte N-grams method; according to the relevancy between the extracted features and a preset mail type, the features are ordered, so that initial feature subsets are generated; according to the Markov blanket approximation algorithm, redundant features in the initial feature subsets are deleted, so that candidate feature subsets are obtained; the candidate feature subsets are predicated through an online logistic regression classifier and evaluated according to predication results, so that the optimal feature subset is selected; according to the selected optimal feature subset, the junk mails are detected through the online logistic regression classifier. By the adoption of the feature selection method and the detection method of the junk mails, the calculation processes for feature selection and detection of the junk mails are simple, time complexity is low, and therefore the junk mail detection accuracy is greatly improved.

Description

technical field [0001] The invention relates to the technical field of computer network security, in particular to a spam feature selection method and a detection method thereof. Background technique [0002] With the rapid development of the Internet, e-mail has become a new type of information transmission tool, and it is widely used in various fields by virtue of its advantages of low price, convenience and speed. Then the widespread use has also brought some negative effects. A large number of spam mails are flooding people's mailboxes, which not only affects the normal use of users, but also damages the image of operators. Many spam systems have emerged as the times require, but they are faced with problems such as large amount of data and low operating efficiency. [0003] Traditional spam filtering methods, many machine learning methods including Flexible Bayes, decision tree, SVM, and Boosting have been applied to spam filtering. From the current research results, ...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to provide a spam feature selection method and its detection method, to solve the existing feature selection method and spam detection method in the existing high computational complexity, time-consuming, and difficult to use in practical applications Problems with good results

Images