| Abstract: |
In recent years, the blockchain has gained widespread attention in secure decentralized computing due to its features of trustless transaction and verifiable computation without relying on a trusted party. But inherent scalability limitations restrict the use of blockchain in high-throughput applications, with leading platforms today capable-to process only 7-30 transactions per second, while enterprises often require thousands. We present SecureChain, a new-architectured sharded blockchain system that is able to deliver unprecedented throughput while preserving security guarantees through three contributions: (1) An adaptive state-sharding mechanism with dynamic rebalancing distributes blockchain state across multiple parallel execution environments, leading to linear scaling of the throughput up to 23,847 transactions per second over the 64 shards; (2) A cross-shard atomic commitment protocol using optimistic execution with zero-knowledge validity proofs maintains integrity while allowing for high parallel processing without incurring significant performance penalties; by factoring out data dependencies and reducing cross-shard transaction latency by 73.2% compared to two-phase committing schemes; and (3) A novel zkSNARK-based verification scheme enables succinct and constant-time verification for arbitrary computations regardless their complexity, reducing the overhead of proof verification by factor 94.7%. Through extensive evaluation on a 1,024-node geographically distributed testbed, we show that SecureChain can achieve high throughput and low latency: for the number of transactions per second (TPS) at up to 23,847 with finality as low as 2.3 seconds; meanwhile it also is resilient to Byzantine faults of up to f < n/3 malicious nodes. Security analysis in the adaptive adversary models verifies they are secure against known attacks such as grinding, long-range and cross-shard double spending attacks. The framework sets new standards for secure distributed co |
