Research On Behavior-based Recognition Of Spam Filtering Techniques

With the widespread use of E-mail, spam filtering technology has become one of the heated point in information filtering research field. The traditional method Intelligent Message Filter works mainly by using the existing text information to filter the content of E-mail, which lead to a large number of computing resources occupy when calculates text content, and also affect the performance of filtration systems.Behavior-based spam filtering technology is the focus of third generation e-mail filtering technology. After excavation, research, summarization on E-mail sending behavior, we give out more comprehensive indicators of behavior filters which not only improve the efficiency but also the accuracy of spam filtering technology ,they are more effective measures for spam filtering.After combination with the artificial immune algorithm, this technology gaines good learning ability and generalization ability, the learning and updating of memory cells makes this system more dynamic and intelligent, it can detect out the unknown spam by antigenic variation and matching, Interaction with user ensure the accuracy of the system to learn.Although content-based filtering having a large consumption of resources defects, many research results have been applied to the field of e-mail filtering as the text information processing technologies has become more sophisticated. According to the deficiencies of Bayesian filtering algorithm in e-mail filtering field, we introduce the minimum risk-weighted to improve the accuracy of Bayesian algorithm and use the sub-word function and word frequency statistics function of Lucene search engine tool to improve the Bayesian algorithm for achieving better results.Content-based filtering tecnology and behavior-based filtering tecnology should not be separated into two branches, they both have characteristics in their own, we have obtained more satisfactory results by a combination of both tecnologies and get more appropriate form of complementary,all above can not only resolve the problem of computing resources occupation, but also the miscarriage of justice in normal e-mail which behavioral characteristics is not obvious during behavior-based filtered .By combining content filtering with behavioral filtering technology in the form of statistical weights, we can easily change the weight of various components under the existing characteristics of spam to enhance the accuracy of the system filter.The test environment is built on a computer network on virtual machine , through mail server communication,we can effectively simulate the real network environment, and start the corresponding experiments.