Byzantine attack detection using optimized dual‐channel capsule generative adversarial network for securing cognitive radio network

Abstract

Cognitive Radio Networks (CRN) is emerged as a solution for the problem of spectrum scarcity as it achieves transmission opportunities in under‐utilized spectrum bands for primary users (PUs). Therefore, in this manuscript, an efficient Byzantine attack detection using optimized Dual‐Channel Capsule Generative Adversarial Network and Auto‐Metric Graph Neural Network (AMGNN) to secure Cognitive Radio Network. Here, two types of layers are utilized that is processing layer and decision layer. In processing layer, the dataset generation based on emission matrix from Hidden Markov Model (HMM) contains labeled data for the legitimate secondary users including malicious Byzantine attackers. In decision layer, the generated dataset is given to Dual‐Channel Capsule Generative Adversarial Network (DCCGAN) to classify secondary users (SUs) as Normal or Attacker. Normally, DCCGAN not exposes any optimization techniques adoption to compute optimum parameters for assuring exact prediction. Thus, capuchin search algorithm (CSA) is employed for optimizing DCCGAN weight parameters. Once the attack is detected, then alert information is fed to AMGNN opportunistic spectrum access. Proposed model executed at MATLAB and performance analyzed with performance metrics like system utility rate Vs available spectrum range, Attack detection rate, Accuracy, Packet Delivery Ratio (PDR), and throughput. The proposed method attains higher accuracy of 96.31%, 97.06%, 93.52%, higher system utility rate of 97.22%, 91.39%, 96.41%, and higher attack detection rate of 96.25%, 94.12%, 97.15% analyzed to the existing models, such as BAD‐AESA‐SCRN, BAD‐BMHHO‐ENN‐SCRN, and BAD‐HCM‐EELM‐SCRN respectively.