Toward Intelligent Intrusion Prediction for Wireless Sensor Networks Using Three-Layer Brain-Like Learning

Abstract

The intrusion prediction for wireless sensor networks (WSNs) is an unresolved problem. Hence, the current intrusion detection schemes cannot provide enough security for WSNs, which poses a number of security challenges in WSNs. In many missioncritical applications, such as battle field, even though the intrusion detection systems (IDSs) without prediction capability could detect the malicious activities afterwards, the damages to the WSNs have been generated and could hardly be restored. In addition, sensor nodes usually are resource constrained, which limits the direct adoption of expensive intrusion prediction algorithm. To address the above challenges, we propose an intelligent intrusion prediction scheme that is able to enforce accurate intrusion prediction. The proposed scheme exploits a novel three-layer brain-like hierarchical learning framework, tailors, and adapts it for WSNs with both performance and security requirements. The implementation system of the proposed scheme is designed based on agent technology. Moreover, an attack experiment is done for getting training and test data set. Experiment results show that the proposed scheme has several advantages in terms of efficiency of implementation and high prediction rate. To our best knowledge, this paper is the first to realize intrusion prediction for WSNs.

Optimization of Intrusion Detection Systems Determined by Ameliorated HNADAM-SGD Algorithm

“…The dependent variable (s) is predicted by a given independent variable (t) to define a linear relationship between (s-input) and (t-output). For linear regression, the hypothesis function is given in Equation ( 14) [20,27] t…”

Application of Machine-Learning Based Prediction Techniques in Wireless Networks

“…Two datasets-one of normal traffic and one of an abnormal attack are used to train the SVM classifier. Since, attacks are continuously evolving, their detection requires the algorithm to learn like the human brain does as detailed in [22]. The authors propose a three-layer hierarchical brain-like learning algorithm for intrusion detection and prediction in wireless sensor networks.…”