Swin-Yoked neural architecture optimized by PaCFIQ for intelligent predictive security in WSN-IoT environments

Abstract

Security in Wireless Sensor Network-based Internet of Things (WSN-IoT) systems is becoming increasingly important in the face of many cyber-attacks. This paper presents a smart framework, Swin-Yoked Neural Approach to Predictive Security (SYNAPSE) with a Parrot-CatchFish Integrated Q-Learning (PaCFIQ) optimizer to address these challenges. The SYNAPSE framework uses Swin Transformer blocks and a yoked neural architecture to improve local-global attention to features and to extract fine-grained attack traces. The PaCFIQ optimizer, drawing inspiration from parrot flight dynamics, fish hunting strategies, and Q-learning-based reinforcement learning, dynamically optimizes hyperparameters and learning rates to enhance convergence. The system starts with sensor data pre-processing and deep Swin feature extraction, yoked neural integration, and adaptive optimization. Evaluation on standard intrusion datasets shows SYNAPSE attains 99% accuracy, 99% precision, 98.9% recall, 98.9% F1-score, and an AUC of 0.99, with an execution time of 10.5 s in contrast to 14.2 s of other models.