Fork-free hybrid consensus with flexible Proof-of-Activity

Highlights

• The first fork-free PoW-based blockchain in the permissionless environment.

• Reducing variance without centralized mining pools.

• Flexible hybrid of PoW and PoS.

Abstract

Bitcoin and its underlying blockchain mechanism have been attracting much attention. One of their core innovations, Proof-of-Work (PoW), is notoriously inefficient which potentially motivates a centralization of hash power, defeating the original goal of decentralization. Proof-of-Stake (PoS) is later proposed to replace PoW. However, both PoW and PoS have different inherent advantages and disadvantages, so does Proof-of-Activity (PoA) of Bentov et al. (SIGMETRICS 2014) which only offers limited hybrids of two mechanisms. On the other hand, the hybrid consensus protocol of Pass and Shi (DISC 2017) aims to improve the efficiency by dynamically maintaining a rotating committee. Yet, there are unsatisfactory issues including chain forks and fair committee election.

In this paper, we firstly devise a generalized variant of PoW. After that, we leverage our generalized PoW to construct a fork-free hybrid consensus protocol. We further combine our fork-free hybrid consensus mechanism with PoS for a flexible version of PoA with tunable parameters between PoW and PoS. Compared with Bentov et al.’s PoA, our “flexible PoA” improves the efficiency, leading to a more applicable consensus protocol.

Introduction

Blockchain, or “Nakamoto chain” (for differentiating it from later proposals), has been attracting much interest (e.g., see [1], [2], [3]) since it first appears as an implicit consensus mechanism used by bitcoin [4] and subsequent decentralized cryptocurrencies (e.g., [5], [6], [7]). Blockchain keeps a growing distributed ledger of blocks, each of which includes an ordered list of transactions. Blockchain is built upon the methodology of Proof-of-Work (PoW) (e.g., see [8]), which requires the creator of a new block to solve a hash puzzle regarding the hash of the previous block, an ordered list of transactions, as well as other necessary information. Solving a hash puzzle regarding some content $w$ is to find a solution $x$ so that $H(x\parallel w)$ falls into a target range. Thereby, any newly generated block is created by an honest node with high probability, as most computing power (called “hash rate”, or “hash power”) solving this puzzle is at hands of honest nodes. After a solution is obtained, the lucky solver (also called miner, for the possibility of gaining some bitcoins after completing this process) can then propose a block containing a list of transactions to the peer-to-peer bitcoin network, and the distributed ledger of blocks grows. PoW ensures that tampering the records on the blockchains requires investing a lot of computing power. We refer this as “traditional PoW”, or just “PoW” when no ambiguity exists.

When multiple new blocks are generated “simultaneously”, the disagreement manifests in the form of a chain fork (or simply a fork) having more than one branch. The fork may be a result of coincidence or tampering attempt from malicious nodes. To confirm which branch is valid, the rule used by the bitcoin system is to pick the first forked branch that is followed by a certain number of blocks. and discard any other branches. As such, honest nodes should only work on the longest valid chain. Resolving the fork tackles the misbehavior of (malicious) miners, i.e., clearing any disagreement and making all nodes concede to “the miner of the next block”. Yet, users have to wait long to make sure one block will not be nullified by other forks. Also, fork leads to issues like selfish mining [9], which undermines both fairness and security. A fork-free blockchain consensus is thus desired.

Serving as a core part of the consensus protocol underlying bitcoin, PoW shows several potential merits such as openness to any participant and good robustness. The puzzle should be hard enough so that expectedly only one block can be solved in a certain period of time, which is ten minutes in bitcoin. PoW-based protocols thus often confirm the validity of a newly added block at an unsatisfactory speed. Since an individual may take years to find a puzzle solution, mining pools emerges which bring us back to a more centralized setting.

Two major approaches are considered for addressing the above issues. The first approach is to replace PoW with Proof-of-Stake (PoS) [10], [11], [12], which moves the decision basis from computing power to possession of stake in the system (e.g., in the form of cryptocurrency). With PoS, specific risk of having a few mining farms dominating PoW is mitigated, and the fork-free property can be achieved. Yet, PoS still faces another kind of centralization risk (from large stakeholders). Another approach is to adjust the protocol of PoW, such as Fruitchain [13] which aims to reduce the variance in mining revenue without a centralized mining pool. Other works are done to provide an instant transaction confirmation [5], [14]. However, to our knowledge, no PoW-based protocol simultaneously achieves the fork-free property, significant improvements to the variance, and instant transaction confirmations. This motivates our work.

We aim to achieve a fork-free property and a smaller variance of miners’ revenues, thereby we change the principle of blockchain mining so that multiple puzzle solutions can be found each round. For the first time, blockchain-based consensus protocol accepts multiple solutions, and we name it “the generalized PoW”. All of these solutions are submitted to a committee directly without causing any fork by the means of a practical Byzantine fault tolerance (PBFT) from the distributed system literature. Moreover, all of them are recorded, so that the history of records is still hard to forge. Based on the idea of hybrid consensus proposed by Pass and Shi [14], and the generalized PoW, we construct a scheme which we call the fork-free hybrid consensus. Note that the protocol of Pass and Shi elects a committee by the blockchain to verify transactions, who are miners of certain blocks. In contrast, our fork-free hybrid consensus protocol lets the committee (instead of block proposers) decide the record for the current round (including transactions, accepted puzzle solutions) and future committee members once for all without any ambiguity.

We can further allow different rules of committee election. Specifically, we establish a function to assign a weight to each candidate according to its PoW power and its PoS capability, and the election is based on such a weight. We thus propose a flexible PoA protocol. This takes a step further from the notion of Proof-of-Activity (PoA) proposed by Bentov et al. [15] which aims to inherit the advantages of both PoW and PoS by determining the miner of a new block by taking into account both its hash power as well as its stake. Basing on the fork-free hybrid consensus, our flexible PoA is also fork-free. Table 1 compares between our constructions and other few consensus schemes. We show the roadmap of our constructions in Fig. 1.

• The first fork-free PoW-based blockchain in the permissionless environment. In bitcoin, the integrity of transactions in a block is guaranteed by fork resolutions (e.g., blocks including double-spending transactions are resolved), since any malicious branch can be outraced by an honest one. We employed the paradigm of hybrid consensus which leverages the security of practical Byzantine fault tolerance (PBFT) to get rid of fork resolution while ensuring transaction integrity. To the best of our knowledge, achieving fork-free property in this way is not yet identified by the literature including the work of Pass and Shi.

• Reducing variance without centralized mining pools. Traditional PoW crucially relies on accepting a single hash puzzle to ensure that existing records cannot be tampered with. Our proposed functionality of generalized PoW accepts multiple solutions for the same puzzle in each round, this reduces the mining-revenues variance. This functionality is hard to realize in bitcoin since its setting provides nothing to “operate” on different solutions. But our fork-free hybrid consensus achieves this functionality by leveraging a rotating committee.

• Flexible hybrid of PoW and PoS. We construct a flexible PoA by having a committee perform the election based on a hybrid weight regarding the participants’ PoW power $w$ and the PoS capability $s$. The relationship between the hybrid weight ($w$ and $s$) can be flexibly determined according to different scenarios. To our knowledge, such a flexibility is never considered in previous works.