Dynamic spam detection system and most relevant features identification using random weight network

Abstract

Nowadays e-mail is being used by millions of people as an effective form of formal or informal communication over the Internet and with this high-speed form of communication there comes a more effective form of threat known as spam. Spam e-mail is often called junk e-mails which are unsolicited and sent in bulk. By these unsolicited emails, the Internet users are hugely impacted in terms of security concerns as well as being exposed to contents that are not appropriate for certain users. There is no way to stop spammers using static filters because almost every other day they find a new way to bypass the filter. New techniques are introduced to elude this system. In this paper, a smart and dynamic(adaptive) system is proposed that will be using Random Weight Network (RWN) to approach spam in a different way and meanwhile this will also detect the most relevant features that will help to design the spam filter. A spam filter with the capability of identifying spam automatically will also be embedded in the proposed system. Also a comparison of different parameters for different RWN models have been shown to determine which model works best with what parameters under different situations.