Intelligent deep federated learning model for enhancing security in internet of things enabled edge computing...

Area of Science:

• Cybersecurity and Privacy
• Artificial Intelligence and Machine Learning
• Internet of Things (IoT) and Edge Computing

Background:

• Rapid advancements in IoT and edge computing introduce significant security and privacy challenges, particularly concerning personal data leakage.
• Centralized Machine Learning (ML) methods struggle with the vast, distributed data generated by IoT devices, necessitating decentralized solutions.
• Federated Learning (FL) offers a privacy-preserving approach, but system heterogeneity in IoT edge environments presents implementation hurdles.

Purpose of the Study:

• To develop an Intelligent Deep Federated Learning Model for Enhancing Security (IDFLM-ES) tailored for IoT-enabled edge-computing environments.
• To address privacy concerns and enhance the security of IoT systems by detecting intrusions.
• To improve the efficiency and convergence of federated learning models in heterogeneous IoT settings.

Main Methods:

• The IDFLM-ES approach utilizes a federated hybrid deep belief network (FHDBN) model with FL on time-series data from IoT edge devices.
• Pre-processing involves data normalization and feature selection using Golden Jackal Optimization (GJO).
• The model learns distributed feature representations to accelerate convergence, with hyperparameters optimized by the Dung Beetle Optimizer (DBO).

Main Results:

• The IDFLM-ES methodology demonstrated superior performance on a benchmark dataset.
• Experimental validation showed an accuracy of 98.24%, outperforming existing models.
• The approach effectively identifies unwanted intrusions, enhancing IoT environment safety.

Conclusions:

• The IDFLM-ES approach provides an effective solution for enhancing security and privacy in IoT edge computing environments.
• Federated learning, combined with deep learning and optimization techniques, can overcome challenges posed by data heterogeneity.
• The proposed model offers a promising direction for secure and privacy-preserving data analysis in distributed IoT systems.