Optimizing QoS and security in agriculture IoT deployments: A bioinspired Q-learning model with customized shards

Area of Science:

• Computer Science
• Artificial Intelligence
• Network Security

Background:

• Agriculture Internet of Things (AIoT) deployments face challenges with existing blockchain-based security and Quality of Service (QoS) models, which suffer from complexity, high latency, and significant energy consumption.
• The scalability of current models is limited by consensus-efficiency and miner-efficiency, hindering real-time performance in large-scale AIoT networks.

Purpose of the Study:

• To design an efficient Q-Learning bioinspired model for enhancing QoS in AIoT deployments using customized shards.
• To address the limitations of existing blockchain models by improving network performance, security, and energy efficiency.

Main Methods:

• A Q-Learning process utilizes continuously updated trust metrics of AIoT nodes to identify suitable miners for block addition.
• A novel Proof-of-Performance (PoP) consensus model with a dynamic consensus function based on miner node performance facilitates block addition.
• Customized shards, configured using Mayfly Optimization (MO) and Bacterial Foraging Optimization (BFO) models, enhance efficiency and scalability.

Main Results:

• The proposed model achieved a 4.5% improvement in mining speed and a 10.4% reduction in energy consumption for mining.
• Throughput during AIoT communications increased by 8.3%, and packet delivery consistency improved by 2.5% compared to existing models.
• The model demonstrated consistent performance even under large-scale attacks.

Conclusions:

• The Q-Learning bioinspired model with customized shards offers a significant improvement in efficiency and performance for AIoT deployments.
• The novel PoP consensus mechanism and optimization techniques effectively address scalability and energy consumption issues in AIoT networks.