Identity Based Attack Detection using Received Signal Strength in MIMO Systems

Abstract

In wireless networks, identity-based attacks are easy to conduct, and as the number of wireless devices grow, these attacks are becoming widespread. Standard cryptographic processes are resource costly and, in some situations, do not provide appropriate protection. Received Signal Strength (RSS) investigations have showed potential in detecting and identifying identity-based attacks. Different RSS-based approaches using SISO systems have been proposed by the researchers in previous work. However, each technique using SISO systems has the drawback that SISO systems donot provide any diversity gain which make it inappropriate for a larger range of wireless networks in a confined setting. In this research work, we propose RSS based MIMO diversity combining techniques for identity-based attack detection. The diversity combining techniques that we consider for identity-based attack detection using MIMO systems are Selection Combining (SC), Equal Gain Combining (EGC), and Maximal Ratio Combining. We first propose to use the spatial correlation of the RSS hereditary from wireless nodes to detect identity-based attacks. In addition, we provide a theoretical explanation of our proposed methods. We then derive the test statistics for detection of identity-based attacks by using unsupervised machine learning algorithm such as K-means and K-medoids algorithm. Simulation results show that our proposed techniques exploiting MIMO diversity combining for identity-based attack detection outperforms SISO systems in terms of false positive rates (FPR) and detection rate (DR). Moreover, it has been found through simulation results that the EGC diversity implemented on spatial correlation using K-means and K-medoids clustering outperform SLC and MRC combining techniques. Secondly, we use supervised machine learning algorithms such as Decision Tree (DT), Logistic Regression (LR), K-nearest neighbors (K-NN), Naïve Bayes (NB), and Support Vector Machine (SVM) for identity-based attack detection using RSS with different MIMO diversity combining techniques. We compare these ML algorithms with MIMO diversity combining techniques using different metrics such as accuracy, precision, DR, and F-measure. Simulation results based on precision, DR, and Fmeasure metric depicts that SVM gives better results as compared to the other ML algorithms.