Spam filter based on geographical location of the sender

Tomáš Caha – Martin Kovařík

   Spam annoys users and poses a security threat. This article proposes a spam filter based on geographical location of the sender determined by IP geolocation. This filter was implemented as a plugin to the SpamAssassin anti-spam software. The plugin allows to define a penalty score for specific countries sending spam. The proposed filter was tested on a dataset of 1500 e-mails consisting of 1200 spam and 300 legitimate e-mails. The Matthews correlation coefficient of the filter has a value of 0.222. This indicates that the proposed spam filter contributes to the correct spam filtering.

Keywords: spam filter, IP geolocation, Matthews corrrelation coefficient

[full-paper]