OAM-GANN: Online Adaptive Memory based Genetically optimized Artificial Neural Network for PUEA Detection in CRN Applications

Abstract

Abstract Cognitive Radio Networks (CRNs) allows opportunistic usage of spectrums owned by licensed users or primary users (PUs). The unlicensed users or secondary users (SUs) that use the spectrum rely opportunistically on spectrum sensing to determine the presence of PU signal. Unfortunately, this attribute opens the door for attacks such as the as the Primary User Emulation Attack (PUEA). This attack happens when an attacker emulates a PU signal. The intention of the attacker might be to grab the vacant channels for its data transmission or entirely disrupt the working of the CRN. Hence it is necessary to combat the consequences of PUEA effectively. Artificial Intelligence (AI) has shown its excellence in various applications including the detection of PUEA. To further enhance the security of the CRN, this research work proposes a novel classification framework called Online Adaptive Memory based Genetically optimized Artificial Neural Network (OAM-GANN) which introduces adaptive online learning of network parameters to identify the presence of PUEA. The proposed OAM-GANN involves Computational Intelligence (CI) algorithm called Memory based Genetic Algorithm (MGA) to optimally tune the hyperparameters of the developed ANN. The advantages of online adaptive training and optimal hyperparameter tuning of the AI model, result in improved security to the network and the data being transmitted in the CRN. The performance of the proposed attack detection model is evaluated in terms of accuracy, sensitivity, specificity, error rate, and detection probability. In addition, the performance of the proposed secure CRN is evaluated in terms of throughput and packet delivery ratio.