Intelligent temporal classification and fuzzy rough set-based feature selection algorithm for intrusion detection system in WSNs

At present, Internet-based information processing systems are challenged by different kinds of threats, which lead to various types of damages that in turn result in significant loss of information in Wireless Sensor Networks (WSNs). Moreover, the stream of data communication through WSNs is continuously increasing in volume. Such huge amounts of data are targeted by attackers frequently, and thus it is crucial to provide high-level security to ensure safe and effective communication of information through the Internet. In order to secure data communication over the Internet through WSNs, Intrusion Detection Systems (IDSs) must be provided as an essential component in networks, databases and cloud systems. Over the previous decade, various approaches were utilized for developing IDSs. Unfortunately, however,none of these systems are completely flawless due to uncertainty surrounding network traffic created by normal users and attackers. Hence, the need to develop efficient IDSs has increased continuously. The present study proposes an adaptive IDS based on Fuzzy Rough sets for attribute selection and Allen's interval algebra, which is applied on network trace datasets in order to select a huge number of attack data for effective prediction of attacks in WSNs. In addition, a fuzzy and rough set based nearest neighborhood algorithm (FRNN) is proposed in this article for effective classification of network trace dataset. This model uses a biased dataset that has 50:50 normal and attack data as opposed to the conventional datasets that have 80:20 normal and attack data. The efficiency of the proposed IDS is increased due to the use of biased data. The combination of feature selection, temporal-based dataset selection, and classification using a biased dataset reduces the false alarm rate and increases detection accuracy. (C) 2019 Elsevier Inc. All rights reserved.