Spam Detection Approaches with Case Study Implementation on Spam Corpora

Spam Detection Approaches with Case Study Implementation on Spam Corpora

Biju Issac (Swinburne University of Technology (Sarawak Campus), Malaysia)
DOI: 10.4018/978-1-60960-015-0.ch012

Abstract

Email has been considered as one of the most efficient and convenient ways of communication since the users of the Internet has increased rapidly. E-mail spam, known as junk e-mail, UBE (unsolicited bulk e-mail) or UCE (unsolicited commercial e-mail), is the act of sending unwanted e-mail messages to e-mail users. Spam is becoming a huge problem to most users since it clutter their mailboxes and waste their time to delete all the spam before reading the legitimate ones. They also cost the user money with dial up connections, waste network bandwidth and disk space and make available harmful and offensive materials. In this chapter, initially we would like to discuss on existing spam technologies and later focus on a case study. Though many anti-spam solutions have been implemented, the Bayesian spam detection approach looks quite promising. A case study for spam detection algorithm is presented and its implementation using Java is discussed, along with its performance test results on two independent spam corpuses – Ling-spam and Enron-spam. We use the Bayesian calculation for single keyword sets and multiple keywords sets, along with its keyword contexts to improve the spam detection and thus to get good accuracy. The use of porter stemmer algorithm is also discussed to stem keywords which can improve spam detection efficiency by reducing keyword searches.
Chapter Preview
Top

Introduction

Over the last years, unsolicited bulk mail, better known as spam, has become one of the most annoying problems of the Internet. The increase of spam emails uses bandwidth and fills up databases and therefore the global network becomes more crowded and less useful. Even though spam emails do not damage the data in the way that viruses do, they do harm the business intentions. For example, spam emails wastes user’s time since the users devoid of anti-spam protection have to check which email is spam manually and then delete it. Sometimes, users can easily overlook or delete important email because of confusing it with spam. Email spamming often contains deceptive, worthless content or even a virus attachment.

Spam emails are getting better in its ability to break anti-spam filters and it would take a great deal of research to get it fully eradicated by coming up with very intelligent anti-spam filters. Spammers are also becoming more innovative, so that the anti-spam research is having a great relevance these days. There are various anti-spam techniques that have been created and implemented since spam started infiltrating user’s inboxes. The most popular and direct way to prevent spam is the anti-spam filters. Anti-spam filters are the software tools that block spam messages automatically. These filters vary in functionality from black list (spammer list) and white list (trusted user list) to content-based filters. There are a lot of anti-spam filters or spam detection schemes available in the market.

The spammer’s methods of avoiding detection evolve constantly, differing significantly from what has been used in the past. For every techniques created for filtering the emails, a new method to spread spam also comes out, making the battle between the spammers and mail agent even more challenging. We would like to introduce a Bayesian approach to the anti-spam solution, considering the context of keywords found. First we implement a simple Bayesian filter based on single keyword sets. Then we improve that by using multiple keyword sets and assigning a higher weightage to them. Finally, we further refine the anti-spam filter by using context matching technique along with the previous steps. The keywords are mapped to a keyword context, which is a collection of other keywords where the specific keyword is found.

The spam relayed by different countries in second quarter of 2007 is shown as a graph in Figure 1 (E-mail spam, na). This gives a good indication that some selected countries are the top relay points of spam emails. The actual spammer may or may not be sending spam emails from the country of his residence or may use compromised PCs elsewhere, even in other countries.

Figure 1.

E-mail spam relayed by country in Quarter 2, 2007

Complete Chapter List

Search this Book:
Reset