Abstract — It has been observed that the underlying reasons for
the continuing growth of the “spam” problem are a lack of reliable
sender authentication and the near-zero cost of sending huge
volumes of marketing material worldwide, via email. Previous
attempts to address these problems either change the fundamental
properties of email, reducing its usefulness to legitimate senders,
or require an infeasible move to new system architectures.
In this paper we present two new techniques for increasing the
level of sender authentication for legacy-system plain text email
addresses. We then show how these Trustworthy Email Addresses
(TEA) can be used in conjunction with a trust and risk-based
security framework as an effective anti-spam tool. Our prototype
Java implementation is then evaluated in the context of a spammer
threat model with an economic analysis of the viability of each
threat.
Index Terms — email spam, computational trust engine,
security cost/benefit analysis, anti-spoofing
I. INTRODUCTION AND PROBLEM OVERVIEW
he worldwide cost of spam has become intolerable [12].
Many efforts have been spent to eradicate spam but none
have, so far, succeeded.
The root cause of spam is ultimately the same property of
email that make it so attractive and useful: the low cost of
communicating with a large number of people all over of the
world. Moreover, the near-zero cost of creating and spoofing
an email identity ensures that even when the sending of
unsolicited bulk messages is prohibited by law or ISP policy,
tracing and punishing the offender is not easy because the
underpinnings of current email systems were not designed with
authorisation and secure authentication in mind. Proposed
solutions which attempt to remedy this oversight have been
dismissed as infeasible in the short term as transitioning all of
the world's email users to a new system is a monumental task
[12, 21].
Authentication systems such as PGP [27] and S/MIME [18]
which are designed to run over top