Title
AntispamLab - A Tool for Realistic Evaluation of Email Spam Filters
Abstract
The existing tools for testing spam fllters evaluate a fllter instance by simply feeding it with a stream of emails, pos- sibly also providing a feedback to the fllter about the cor- rectness of the detection. In such a scenario the evaluated fllter is disconnected from the network of email servers, fll- ters, and users, which makes the approach inappropriate for testing many of the fllters that exploit some of the informa- tion about spam bulkiness, users' actions and social relations among the users. Corresponding evaluation results might be wrong, because the information that is normally used by the fllter is missing, incomplete or inappropriate. In this paper we present a tool to test spam fllters in a very realistic scenario. Our tool consists of a set of Python scripts for Unix/Linux. The tool takes as inputs the fllter to be tested and an afiordable set of interconnected machines (e.g., PlanetLab machines, or locally created virtual ma- chines). When started from a central place, the tool uses the provided machines to build a network of real email servers, installs instances of the fllter, deploys and runs simulated email users and spammers, and computes the detection re- sults statistic. Email servers are implemented using Postflx, a standard Linux email server. Only per-email-server fll- ters are currently supported; testing per-email-client fllters would require additional development of the tool. The size of the created emailing network is constrained only by the number of available PlanetLab or virtual machines. The run time is much shorter then the simulated system time, due to a time scaling mechanism. Testing a new fllter is as sim- ple as installing one copy of it in a real emailing network, which unifles the jobs of a new fllter development, testing and prototyping. As an example of how to use the tool, we test the SpamAssassin fllter.
Year
Venue
Keywords
2007
CEAS
virtual machine,testing,spam,evaluation,filter,social relation
Field
DocType
Citations 
World Wide Web,PlanetLab,Virtual machine,Computer science,Server,Correctness,Unix,Email spam,Python (programming language),Scripting language
Conference
3
PageRank 
References 
Authors
0.48
3
4
Name
Order
Citations
PageRank
Slavisa Sarafijanovic1958.20
Luis Hernandez230.82
Raphael Naefen330.48
Jean-Yves Le Boudec45075471.48