Federated reinforcement learning-driven multi-task optimization for robust and ethical edge internet of things...

Area of Science:

• Computer Science
• Cybersecurity
• Artificial Intelligence

Background:

• Edge Internet of Things (IoT) systems for campus security face challenges balancing detection accuracy, resource efficiency, privacy, and adversarial robustness.
• Traditional Federated Learning (FL) and quantization methods are insufficient for heterogeneous devices and ethical constraints.

Purpose of the Study:

• To propose a Federated Reinforcement Learning (FRL)-driven multi-task collaborative optimization framework for secure and efficient edge IoT environments.
• To develop a hierarchical architecture for dynamic adaptation of parameters and ensure continuous ethical compliance through distributed auditing and automated constraint enforcement.

Main Methods:

• A novel FRL-driven framework with a perception-decision-constraint architecture.
• Dynamic adaptation of quantization precision, defense intensity, and privacy parameters.
• Closed-loop optimization integrating technical robustness and normative governance.

Main Results:

• Achieved 94.3% intrusion detection accuracy and 89.3% F1-score for User-to-Root (U2R) attacks on the NSL-KDD benchmark.
• Reduced energy consumption by 66.1% and latency by 72.8%.
• Limited adversarial attack success rate to 23.4% and privacy leakage incidents to 2.4 per month under Differential Privacy (DP) constraints (ε=2.0).

Conclusions:

• The proposed FRL framework offers an integrated, scalable, and ethically accountable solution for campus network security.
• Demonstrates FRL's potential for synergistic optimization across performance, robustness, and compliance in edge intelligent systems.