Smart and Secure Anomaly Detection in IoT Using Quantum Inspired Kernel Ensemble Learning

Abstract

The exponential growth of the Internet of Things networks has greatly increased the attack surface for cyber threats, necessitating the use of strong and perceptive anomaly detection systems. Traditional intrusion detection systems often struggle with high false alarm rates, limited generalization, and inefficiency in processing high-dimensional heterogeneous IoT data. To address this issue, a novel Hybrid Quantum-Enhanced Kernel Ensemble Learning model is proposed for efficient and scalable anomaly detection in the Internet of Things environment. Initially, different datasets were gathered and preprocessed using data encoding, data cleaning, missing data handling, and null value removal. A novel Crossover Strategy Enhanced Wombat Optimization Algorithm is developed to select optimal features that are suitable for classifying anomalies. To improve detection performance and lower computational overhead, the most pertinent features are chosen using a Gradient-based Feature Evaluator combined with a Quantum Kernel Estimation Layer. In order to attain strong performance, the refined features are subsequently fed into an ensemble classification framework that integrates predictions. Thus, the proposed framework guarantees safe, scalable, and intelligent anomaly detection designed for real-time Internet of Things network protection, while also reducing false positives.