On extendable software architecture for spam email filtering

The research community and the IT industry have invested significant effort in fighting spam emails. There are many spam email filters available and also in operation. Yet we are still inundated with spam emails everyday. This is not because the filters are not powerful enough, but because the filtering systems are not flexible enough to adapt the new development of spam techniques, such as HTML tagging, image based spam, and keyword obfuscating etc. In this paper, we propose to use dynamic multiple normalizers as the preprocessors for spam filters. The normalizers convert an email to its plain text format, called normalization. With the help of the normalizers, spam filters only need to deal with plain text format, which is what the filters are good at.