LightDew: Lightweight Blockchain Assisted Dew Computing Framework for Smart Assisted Living
Sudip Chatterjee, Pronaya Bhattacharya, Debashis De · IEEE Transactions on Consumer Electronics · 2024
The paper proposes a framework LightDew, which leverages the Dew Computing (DC) paradigm in Smart Assisted Living (SAL) ecosystems. DC allows effective communication, less resource utilization, and real-time decision making capability. However, due to the open nature of communication channels, security and trust among data transmitted over autonomous networks is important. Firstly, at dew layer, we propose a shallow neural network model, which is named as Lightweight Activity Recognition (LAR) model. The LAR outputs (local predictions) are securely signed and hashed using the lightweight secp256k1 curve, which assures data integrity and authentication. The hashed reference is stored as a transaction in a local blockchain node, and the entire data is stored in an attached offline Interplanetary File Systems (IPFS) ledger. For Human Activity Recognition (HAR), we have considered two benchmarking datasets, the WISDM dataset, and the E-care Home dataset. We observed an average training accuracy of 0.895, validation accuracy of 0.87. The average response time significantly improved by 32.65%. The signing time is 10.21 ms using the elliptic sec256k1 curve, mining time of a block is 13.24 seconds, and total energy consumption is 11.88 kWh, which makes the framework suitable for real-time operations.