A Plug-and-Play Long-Range Defense System for Proof-of-Stake Blockchains

【Author】 Ng, Lucien K. L.; Chatzigiannis, Panagiotis; Le, Duc V.; Minaei, Mohsen; Kumaresan, Ranjit; Zamani, Mahdi

【Source】COMPUTER SECURITY-ESORICS 2024, PT IV

【影响因子】

【Abstract】In recent years, many blockchain systems have progressively transitioned to proof-of-stake (PoS) consensus algorithms. These algorithms are not only more energy efficient than proof-of-work but are also well-studied and widely accepted within the community. However, PoS systems are susceptible to a particularly powerful "long-range" attack, where an adversary can corrupt the validator set retroactively and present forked versions of the blockchain. These versions would still be acceptable to clients, thereby creating the potential for double-spending. Several methods and research efforts have proposed countermeasures against such attacks. Still, they often necessitate modifications to the underlying blockchain, introduce heavy assumptions such as centralized entities, or prove inefficient for securely bootstrapping light clients. In this work, we propose a method of defending against these attacks with the aid of external servers running our protocol. Our method does not require any soft or hard-forks on the underlying blockchain and operates under reasonable assumptions, specifically the requirement of at least one honest server. Central to our approach is a new primitive called "Insertable Proof of Sequential Work" (InPoSW). Traditional PoSW ensures that a server performs computational tasks that cannot be parallelized and require a minimum execution time, effectively timestamping the input data. InPoSW additionally allows the prover to "insert" new data into an ongoing InPoSW instance. This primitive can be of independent interest for other timestamp applications. Compared to naively adopting prior PoSW schemes for InPoSW, our construction achieves >22x storage reduction on the server side and >17900x communication cost reduction for each verification.

【Keywords】

【发表时间】2024

【收录时间】2024-11-27

【文献类型】

【主题类别】

--