Fraud and Data Availability Proofs: Maximising Light Client Security and Scaling Blockchains with Dishonest Majorities

This paper explores the critical roles of fraud proofs and data availability proofs in improving the security of light clients and enabling scalable blockchain architectures, particularly those employing sharding. The authors demonstrate how light clients, which only download block headers, can achieve near full-node security by relying on fraud proofs to detect invalid state transitions or blocks. Crucially, they introduce the concept of data availability sampling (DAS), where light clients probabilistically sample small portions of block data to ensure that the entire block's data is indeed available to the network for reconstruction, even if it's not downloaded by the light client itself. This mechanism is vital for sharded blockchains, where a malicious shard might attempt to produce an invalid or unavailable block without detection. The research shows how these proofs can secure a sharded system even under a dishonest majority of participants, enhancing robustness and allowing for significant scaling. This work forms a cornerstone for the design of modern modular blockchains and light client security models.