Establishing Trust Between Mail Servers to Improve Spam Filtering

This paper proposes a new way to improve spam filtering based on the establishment and maintenance of trust between mail domains. An architecture is presented where each mail domain has an associated trust manager that dynamically records trust measures pertaining to other domains. Trust by one mail domain in another is influenced by direct experience as well as recommendations issued by collaborators. Each trust manager interacts with local spam filtering and peer trust managers to continuously update trust. These trust measures are used to tune filter sensitivity. A simulation set-up is described with multiple nodes that send and receive mail, some of which is spam. Rogue mail servers that produce spam are also introduced. Results of these simulations demonstrate the potential of trust based spam filtering, and are assessed in terms of improvements in rates of false positives and false negatives.