• Lars Tabro Sørensen
  • Martin Møller Larsen
2. term, Computer Science, Master (Master Programme)
In modern world spam has become a great threat for the entire email system. Advanced techniques are constantly being developed by spammers to bypass anti spam tools. This article aims to describe how to counter three of these techniques, namely unicode, scrambling and misspell obfuscation. We introduce WordAdjust - a frontend deobfuscation filter for the state of the art anti spam tool SpamAssassin. This deobfuscation filter is capable of deobfuscating 3 types of obfuscation methods; unicode, scrambling and intentional misspell. The Misspell deobfuscation is based on N-gram fuzzy search techniques. SpamAssassin has a hard time dealing with spam emails that are obfuscated by these 3 methods. Experiments shows how WordAdjust increases the average SpamAssassin score by 56\% on intentional obfuscated spam emails. Further experiments shows how WordAdjust enables SpamAssasin to catch, in average, 10\% of the spam mails that normally would end in the users inbox.
Publication date2008
Number of pages21
Publishing institutionInstitut for Datalogi
ID: 14459757