SenFEED: Dynamic Decentralized Oracle Services for Accurate and Real-Time Sensor Data

Smart contracts, as self-executing programs on blockchains, have been applied in many Internet of Things (IoT) systems to achieve decentralized, fair, and auditable transaction services. However, since smart contracts cannot natively fetch real-world data, this limitation hinders the full economic potential of smart contracts in IoT. A promising solution is to leverage oracle services (also known as data feeds), which enable smart contracts to access and interact with external data sources. Nonetheless, these services must contend with the streaming nature of sensor data in IoT and the dynamic joining or leaving of service nodes in the blockchain. To address these challenges, we propose dynamic decentralized oracle services called SenFEED. Specifically, we integrate iterative-based and incremental-based truth discovery algorithms to achieve accurate and real-time truth inference on IoT streaming data. Furthermore, to address the challenge of reaching consensus in a decentralized setting, we design a new Byzantine fault-tolerant protocol that allows service nodes to join or leave without requiring a system reset. Theoretical analysis and extensive experiments confirm the effectiveness of SenFEED, demonstrating significant improvements in scenarios with a larger number of nodes and robust support for data streams in IoT systems.