Spoofing Attack Detection Using Receive Signal Strength in Massive MIMO System

Abstract

With exponential growth in the public demand for higher data rates, wireless networks are expanding day by day, as more devices are connecting to the networks. To meet the demands for higher data rates, the researchers are trying to optimally utilize the available resources. One such technology which is making its way into future wireless systems is massive (large-scale) multiple input multiple output (mMIMO), which aims to provide high data rates as per the public demands. However, as wireless networks grow wider, it is becoming more prone to identity-based attacks, as it is easy to conduct spoofing attacks within organizations that are connected wirelessly. As a countermeasure, standard cryptographic techniques can be exploited, however, they require more resources as well as are not able to provide security in all cases. Besides, there are attacks that can be initiated within the organization. Moreover, identity-based attacks (IBA) become even more challenging when the spoofer/ eves dropper is changing its transmit power. This research work addresses IBA detection in mMIMO systems when the eve’s dropper is changing its transmit power. Our proposed techniques for IBAs in mMIMO systems employ MIMO diversity combining techniques based on the received signal strength (RSS), where the spoofer is considered to be transmitting at different power. The receiver-combining techniques considered in this research work are, maximal ratio combining (MRC), selection combining (SLC), and equal gain combining (EGC). Moreover, it has been found in the literature that two factors, i.e., distance and noise, are being considered mostly to achieve the desired results. However, to achieve more accurate results, we consider that Eve’s dropper in varying its transmit power along with the distance and noise factors. As the attacker is considered to be transmitting at varying power, this makes the proposed methodology more appropriate and challenging for attack detection. Firstly, exploiting the spatial correlation, analytical expressions for IBA detection using different combining diversity techniques are calculated. Simulation results are then used to validate the analytical expressions. Secondly, k-means and k-mediods unsupervised machine learning (ML) algorithms are used with different MIMO combining techniques to improve the performance of the system in terms of IBA detections. Simulation results presented show that by exploiting the MIMO combining diversity techniques for IBA detection in mMIMO systems VI using unsupervised ML techniques, the performance of the systems can be improved in terms of the detection rate (DR) as well as the false positive rate (FPR). As a further part of the thesis, we propose supervised ML algorithms with MIMO combining diversity techniques for IBA detection in mMIMO systems. The ML algorithms used in this thesis are k-nearest neighbor (kNN), Naive Bayes (NB), and support vector machine (SVM). Simula-tion results show improvement in the DR and PFR using supervised machine learning techniques as compared to the unsupervised ML algorithms. Moreover, it has been found through simulation that SVM with SLC has the best performance out of all the proposed benchmark algorithms.