A federated edge intelligence framework with trust based access control for secure and privacy preserving IoT systems

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Internet of Things

Background:

• Rapid growth of IoT ecosystems presents significant security and privacy challenges.
• Traditional cloud-centric security models are inadequate for IoT due to bandwidth, latency, and trust requirements.
• Edge computing offers a decentralized approach to address IoT security limitations.

Purpose of the Study:

• To propose the Artificial Intelligence-Driven Secure Edge Trust Framework (AI-SET) for enhanced IoT security.
• To integrate network intrusion detection with federated learning for adaptive trust-based access control.
• To provide a resilient and privacy-preserving security solution for IoT systems.

Main Methods:

• Implemented an Edge-Resident Intrusion Detection System (IDS) with lightweight AI algorithms for real-time anomaly detection.
• Utilized privacy-preserving federated learning with a modified FedAvg algorithm, incorporating differential privacy and homomorphic encryption.
• Developed a dynamic access control system employing trust assessment models for real-time device permission evaluation.

Main Results:

• AI-SET demonstrated higher accuracy in intrusion detection and improved communication performance compared to standard methods.
• The framework achieved superior access control security, proving immunity against model poisoning and system breaches.
• AI-SET maintained low operational costs and ensured secure data privacy throughout its operation.

Conclusions:

• AI-SET offers an adaptable, resilient, and privacy-conscious security framework for future IoT deployments.
• The framework effectively combines edge intelligence, secure network operations, and automated trust management.
• Edge-based AI and federated learning are crucial for addressing complex IoT security challenges.