An enhanced approach towards privacy preserving collaborative spam detection using email layout abstraction

Abstract

Spam is generally an unsolicited electronic junk mail that includes text messages, images or videos and is sent without the consent of the recipients. Spam messages are broadcasted to large number of email users occupying larger bandwidth. It is not only obstructing the network traffic but also forms a base for email viruses and denial of service attacks. Moreover, spam messages contain mostly offensive and fraudulent texts that are unpleasant to the recipients. Email users are drowned with nearly 50% of spam messages with new content and new addresses in their inbox daily. Spam messages may destroy email servers with potentially harmful information and the users need to spend certain amount of time to identify and analyze spam messages in their inbox and delete them. Several spam filtering techniques are proposed to identify solicited and unsolicited messages; however, email spammers use dynamic new structures to thwart all the techniques and conceal email content. The main problem that arises with spam is that spammers devise new ways to attack the spam filters and thereby benefit from sending large amount of spams. The primary challenge is to develop a system that can deal with newly arising spams. newlineExisting spam filtering techniques are either list-based filters or content-based filters or a collaborative response system which generally identify duplicate contents, fraudulent texts or the disreputable servers. Though these filters provide better accuracy rate in spam classification, they are prone to erroneous misclassification of hams as spams. This research work proposes a collaborative Spam detection system that uses email layouts and fingerprints to identify spam messages. Collaborative approach collects the feedback from the users regarding what mails are spams and consequently develops a model against it. The incoming emails are mapped to a known iv newlinespam database using near duplicate matching scheme. Overall three key processes are involved in this spam detection approach. First, a layout abstraction