Deobfuscation based on edit distance algorithm for spam filitering

Spamming problem has been grown rapidly in the Internet. An adversary obfuscates the spam message by misspelling or inserting useless characters to mislead the decision of the spam filter. Humans still can understand the original meaning of the camouflaged words but the spam filter cannot recognize them. This paper focuses on the well-known obfuscation problem which uses non-alphabetical characters, e.g. Viagra is modified to V!@gr@. The string edit distance algorithm is revised for handling the non-alphabetical characters. The proposed deobfuscation method outperforms than the traditional string edit distance algorithm in the experiment.