MQDEB-HTGRU: A Machine Learning Methodology for Measuring the Quality of Untrusted Off-Chain Data Entering Blockchain

Smart contracts (SC) represent business logic code in blockchain. They are unable to access data from off-chain sources on their own. Instead, SC communicates with oracles; off-chain data sources provide data feeds to SC. However, there is a risk of oracles providing malicious or corrupted data, which leads to inaccurate outcomes. In this research, we develop a machine learning methodology that measures the trustworthiness degree of data coming from oracles to blockchain. The proposed system consists of a deep learning model called Hybrid Transformer Gated Recurrent Unit (HTGRU), which can predict cryptocurrency prices as an example of Oracle’s data, and SC which requests needed data from oracles and receives the predicted data value from HTGRU and calculates the trustworthiness degree of the requested data. Results show the proposed methodology has the best-performing model for Bitcoin, Ethereum, and Monero price prediction, with RMSEs of 0.01165, 0.01247 and 0.01040 respectively when compared with previous ML techniques.