A proposal of a method to detect spam from information messages

Abstract

This paper presents a spam detection algorithm that uses solely traffic flow logs in the form of Netflow messages. Internet service providers must detect spam in order for their entire subnets not to be marked as spamming stations. The algorithm was drafted based on an analysis of various datasets containing Netflow records. These datasets consist of valid e-mails, spam and common non e-mail related traffic. The algorithm uses domain name system blacklist verification as the first step of identifying a spamming station. Furthermore, theoretical models of valid clients and spammers have been laid out. In continuation of this work, the dataset will be studied to find correlation with the models. Included in the tracked parameters one can find the number of incoming and outcoming messages, timestamps amongst others.