Abstract:

Current spam filtering algorithms based on artificial immune system consider little about the noise word attack, so an immunebased antinoise word attack spam filtering model, named ANWAIS, is proposed in order to solve the problem. The algorithm uses the Mutual Information Difference as the Evaluation function to discard the good word in the spam and the spam word in the normal email during the stage of the generation of the gene library, so that the gene library can better reflect the characteristics of spam emails. Meanwhile, it can guarantee the purity of the gene library through maintaining the discard word table during the stage of the updating of the antibody. Experimental results show that ANWAIS can obtain higher quality antibody and have better classification performance than that of other spam filtering algorithms without considering the noise word attack.

Key words: artificial immune;noise word attack;spam filter;mutual information difference;gene library