In this paper, we propose a new rule based attribute selection algorithm for removing the redundant attributes which are used in decision making on intrusions in wireless sensor networks. This work focuses mainly on finding important attributes to find Denial of Service attacks. In addition, we used an enhanced MSVM classification algorithm that was developed by extending the existing MSVM algorithm. The experimental results show that the proposed methods provide high detection rates and reduce false alarm rate. This system has been tested using KJDD'99 Cup data set. © 2012 Published by Elsevier Ltd.