Spam filtering using Kolmogorov complexity analysis

One of the most irrelevant side effects of e-commerce technology is the development of spamming as an e-marketing technique. Spam e-mails (or unsolicited commercial e-mails) induce a burden for everybody having an electronic mailbox: detecting and filtering spam is then a challenging task and a lot of approaches have been developed to identify spam before it is posted in the end user's mailbox. In this paper, we focus on a relatively new approach whose foundations rely on the works of A. Kolmogorov. The main idea is to give a formal meaning to the notion of 'information content' and to provide a measure of this content. Using such a quantitative approach, it becomes possible to define a distance, which is a major tool for classification purposes. To validate our approach, we proceed in two steps: first, we use the classical compression distance over a mix of spam and legitimate e-mails to check out if they can be properly clustered without any Supervision. It has been the case to highlight a kind of underlying structure for spam e-mails. In the second step, we have implemented a k-nearest neighbours algorithm providing 85% as accuracy rate. Coupled with other anti-spam techniques, compression-based methods could bring a great help in the spam filtering challenge.