DiffMail: A Differentiated Message Delivery Architecture to Control Spam

Unsolicited bulk electronic mail (spam) is increasingly plaguing the Internet Email users and deteriorating the value of Email as a convenient communication tool. Although many different spam control schemes have been proposed and deployed on the Internet, the proportion of Email spam seen on the Internet has been increasing in recent years. In this paper we argue that the difficulties in controlling spam can be attributed to the lack of user control on how different Email messages should be delievered on the Internet. In the current Email delivery architecture, a party can at will force another party to be involved in an Email communication, regardless of whether the latter is willing to accept the message. Based on this observation, we propose a differentiated message delivery architecture—DiffMail. In DiffMail, a user can classify Email senders into multiple classes and handle messages from each class differently. For example, the user may directly accept messages from regular contacts, while asking other senders to hold their messages at their own mail servers. Such messages can be retrieved from the sender’s mail server if and when the receiver wishes to do so. In this way, DiffMail achieves several appealing objectives. Amongst others, regular correspondence is handled in the same way as in the current Email architecture, spammers are discouraged from blindly sending spam to arbitrary users, and it helps to improve the effectiveness of real-time blacklists of spammers. In this paper we present a detailed design of the DiffMail architecture and conduct empirical studies to illustrate its performance trade-offs using real-world Email archives. Furthermore, we also illustrate how DiffMail can be incrementally deployed on the Internet.