Thanks to built-in immutability and persistence, the blockchain is often seen as a promising technology to certify information. However, when the information does not originate from the blockchain itself, its correctness cannot be taken for granted. To address this limitation, blockchain oracles -- services that validate external information before storing it in a blockchain -- were introduced. In particular, when the validation cannot be automated, oracles rely on humans that collaboratively cross-check external information. In this paper, we present DeepThought, a distributed human-based oracle that combines voting and reputation schemes. An empirical evaluation compares DeepThought with a state-of-the-art solution and shows that our approach achieves greater resistance to voters corruptions in different configurations.
View on arXiv