Ройовий інтелект для IoT: оптимізований алгоритм і гібридна стратегія розгортання для виявлення вторгнень

The proliferation of Internet of Things (IoT) systems has led to the creation of large, heterogeneous, and noisy distributed sensor data streams. Conventional machine learning techniques (e.g., support vector machines, k-Nearest Neighbors) are usually ineffective in these applications as they require feature extraction and may not be suitable for the data pattern changing over time. Swarm Intelligence (SI) based algorithms, motivated by the collective behavior of biological entities provides a strong alternative by decentralized, adaptive and global optimization in parallel. This paper presents an end-to-end SWARM Intelligence based IoT intrusion detection framework consisting of advanced Particle Swarm Optimization (PSO) algorithm with adaptive inertia weight and a pragmatic hybrid edge-cloud deployment scheme. On a public dataset MQTT-IoT-IDS2020, our model has 96.8% classification accuracy which is 7 – 12% higher than most of the traditional machine learning classifiers and also outperforms the best-reported accuracy from existing works. Inference time. The final optimized model has an average inference time of 4.8 ms on edge-like hardware, proving its adequacy for real-time use case. Experimental tests confirm the performance, robustness and efficiency of the proposed SI framework for collections IoT in motion combining performance with constraints.