Proof-of-Work (PoW)¹²

As the fundamental underpinning of the Bitcoin Blockchain developed by Satoshi Nakamoto, PoW is the most frequently used and well-known consensus mechanism. PoW involves “nodes” solving complicated, asymmetrical mathematical puzzles to produce new blocks in a process known as “mining”—thereby showing proof of work. The Bitcoin protocol adjusts the difficulty of these puzzles to ensure that a new block is produced every 10 minutes. Although the puzzles are designed to be hard to solve, they are easy for the network to verify. Sometimes, several nodes may solve the puzzle, which can create parallel blocks, more commonly known as “forking.”¹³ In these instances, the forks are usually only temporary, as all nodes will eventually move to the longest chain while the other chains get discarded by the protocol.

PoW uses two reward mechanisms to incentivize nodes to be active on the network and ensure a large and diverse number of nodes. On the Bitcoin Blockchain, nodes that solve the puzzle get to add their block to the blockchain and are rewarded with new Bitcoins. These rewards are halved every four years and from May 2020, the block reward fell from 12.5 Bitcoins to 6.25 Bitcoins. This concept of rewarding active nodes with crypto assets is replicated in many other consensus mechanisms. The second reward is the transaction fee. When users send Bitcoins, they attach a transaction fee and the higher the fee, the more likely a node will validate the transaction—meaning a faster transaction time. PoW mechanisms allow for large number of nodes to participate in the network, which can make the network scalable. The larger the network of nodes, the higher the hashrate, the less likely it is for a single node to hold power over the network and engage in fraudulent transactions. This approach ensures that the mechanism is sufficiently robust to ensure network security. From a regulatory perspective, nevertheless, the lack of control in terms of who can enter and participate in the system can be a problem.

To solve the mathematical puzzles generated by the Bitcoin protocol, nodes need to use “brute force,” which, in turn, consumes considerable energy because brute force requires specialized computing systems to run through all possible solutions until the winning solution is found—an effort that uses significant power. The IMF has set out that the financial sector has an important role to play in the fight against climate change and seeks to support reductions in climate change risk and mitigating the impact of adverse climate events.¹⁴ However, the total energy usage of Bitcoin mining is comparable to Poland at 140 Terrawatt-hour, and so runs counter to this aim.¹⁵ Regulators using or supporting consensus mechanisms that rely on large-scale energy use should pay attention to this energy consumption; most of them will likely find such damaging impacts to the environment unacceptable.

Although PoW guarantees eventual consistency in the blockchain, there may be some instances where there are competing forks, which can impact the network by making it slow, expensive, and inefficient. Forks can lead to slower settlement times, making the usage of PoW inefficient in certain areas of financial services (for example, payments). Currently, the Bitcoin Blockchain can process approximately 7 transactions per second, and although the development of the Lightning Network offers promise in this area (albeit with its own flaws), its transaction rate is currently much lower than that of traditional payment mechanisms like Visa—which averages roughly 1,700 transactions per second.

Inconsistencies can also lead to risks to settlement finality. Transactions between counterparties carry risks—including credit, liquidity, operational, and legal risks—all of which can trigger systemic hazards. Rules around settlement finality aim to mitigate these risks; however, the possibility of forks in a blockchain makes achieving settlement finality difficult. This probabilistic settlement creates challenges around ensuring PoW-based blockchains fit within broader settlement regulation, as well as the impact this has on areas like custody. The proposed European Union DLT Sandbox aims to identify—among other things—where such legal issues could interfere with existing regulatory frameworks.

Further, PoW is seen to be potentially centralizing, which could negate the security that PoW offers. As mathematical puzzles become increasingly complex, more powerful computing power is needed to solve them. Given the large cost involved, such technology is available only to certain individuals or entities, or where mining pools are formed. This scenario can give rise to questions around the powers, tools, and options available for the regulator to fix an eventual market failure (for example, an individual who, or entity that, becomes “too big to fail”).

Authorities should note that although PoW consensus mechanisms lay the foundation for secure and resilient blockchain systems, they can be slow, they consume significant energy, and settlement is probabilistic. PoW is one of the most secure validation methods and can be useful in records and supply chain management, but the attendant lack of speed, scalability, and settlement finality—and PoW’s considerable energy consumption—are likely to pose regulatory risks that many authorities will be unwilling or unable to accommodate in areas like payments. DLT based on PoW consensus mechanisms might not ft within existing regulatory frameworks focused on settlement finality. Supervisors should also be aware that many blockchains that make use of PoW tend to be decentralized, where not all identities of nodes are known, and so can give rise to supervisory difficulties.

Bitcoin Mining Pools Janaury 2021 to January 2022¹⁶

Citation: FinTech Notes 2022, 003; 10.5089/9781616358280.063.A001

Source: One-year pool distribution calculated by blocks from BTC.com.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Bitcoin Mining Pools Janaury 2021 to January 2022¹⁶

Citation: FinTech Notes 2022, 003; 10.5089/9781616358280.063.A001

Source: One-year pool distribution calculated by blocks from BTC.com.

• Download Figure
• Download figure as PowerPoint slide

Bitcoin Mining Pools Janaury 2021 to January 2022¹⁶

Citation: FinTech Notes 2022, 003; 10.5089/9781616358280.063.A001

Source: One-year pool distribution calculated by blocks from BTC.com.

• Download Figure
• Download figure as PowerPoint slide

Proof-of-Stake (PoS)¹⁷

In a PoS consensus mechanism, an algorithm randomly selects validators for block creation based on the amount that token holders stake from their crypto-asset ownership. The first step is selecting a proposer, then a proposed block, and then validation of the proposed block. Holders with larger ownership of the native token have a greater chance of being selected—like playing the lottery. Even though everyone who buys a ticket has a chance of winning and selection is random, those with the most tickets have the greatest odds of winning.

By not requiring energy-intensive mining operations, PoS improves on some of the weaknesses in PoW consensus mechanisms, such as large energy consumption, while preserving network security. This lessened energy consumption also means a lower need to issue many new coins to incentivize nodes to participate in the network. It also limits the risks of a 51 percent attack: although it would be very difficult and expensive for anyone to carry out a successful 51 percent attack in a large PoW-based blockchain, it is even more expensive to do it in a large PoS-based one.

Given that nodes aren’t working to solve complex mathematical problems and there are fewer validating nodes, transaction rates can be increased—but settlement issues remain. A greater transaction throughput makes the PoS model much more useful in certain financial services contexts, like facilitating payments; however, to increase the chances of being selected, validating nodes might vote on multiple blocks—even those whose underlying information might be incorrect, creating risks around broader market integrity. PoS models can do this because, unlike PoW models, validating nodes in PoS don’t have to expend anything (like energy). While voting on multiple blocks maximizes the chances of nodes receiving a reward through transaction fees, it also increases the risks of multiple forks, which can create uncertainty with settlement finality; this situation is known as the “Nothing at Stake” problem. Newer models of PoS seek to solve it by creating monetary penalties for the work validators do on blocks that do not get included in the chain. These models, however, are still in their infancy.

Because ownership directly correlates with the chances of being selected, PoS consensus mechanisms can theoretically create a community where richer individuals or entities are more likely to be selected. This means those participants are also more likely to be rewarded, fueling an environment where the rich get richer. This scenario can lead to those participants with smaller holdings exiting the network if they aren’t able to generate rewards. In effect, PoS consensus mechanisms could create conditions where the network isn’t inclusive, as certain members (that is, those with larger holdings) are more likely be favored over others, increasing the potential for centralization.

Weaknesses within PoS consensus mechanisms might impact stability and integrity, such as centralization in smaller networks and the inefficiency of staking. In the long term, these flaws can lead to centralization-related issues, which can be particularly problematic in smaller networks or networks in their infancy—and can therefore impact market integrity. Authorities could consider elements like sandboxing to help protect nascent networks.¹⁸

The potential for centralization is similar for PoW, where participants that can afford the most powerful machines are better able to generate rewards; however, PoW-based advantages in terms of centralization are based on the different nature of the capital used for validation and the friction to switch between the currency and the hardware and electricity. Even in larger PoS networks, the potential for validator cartels to form can lead to concerns around centralization, while exchanges and wallet providers could also theoretically exercise disproportionate control given their large holdings. PoS is also inefficient in its use of network-native resources. Given that crypto assets might be locked up for staking purposes, PoS removes the ability to transfer or spend a proportion of the total number of crypto assets in circulation. A liquidity shortage could arise if token holders hoard their tokens to increase their chance of being selected as validators; this act would lower the speed of transaction rates, and the network could suffer from the lack of circulation.

Safeguarding the integrity of financial systems is an important aspect of the BFA, so regulators should think about appropriate network security and fairness. Authorities should consider those frameworks that incorporate appropriate systems and controls to mitigate against cyber and operational risks, such as the Basel Committee on Banking Supervision (BCBS) Principles for Operational Resilience (BCBS 2021). In line with the BFA elements on improving financial inclusion and ensuring open, free, and contestable markets, authorities should also consider how to work with market participants developing these networks to ensure that they are as inclusive and collaborative as possible and that these networks do not work to benefit a small privileged group and lead to areas of financial markets that lack contestability. Finally, due to the Nothing at Stake problem, authorities may want to pay attention to the risks created by competing chains.

Delegated Proof-of-Stake (DPoS)¹⁹

DPoS adds a democratic element to the PoS consensus mechanism by outsourcing the validation process. As with PoS, block validation is randomized, and individuals or entities (stakeholders) who stake the greatest amount of a particular crypto asset are more likely to win the opportunity to validate a block and generate a reward. The DPoS model operates a voting system where the stakeholders chosen to validate a block can outsource their work to a third party. These third parties are known as “witnesses” and are responsible for achieving consensus during the generation and validation of new blocks. Rewards are shared between the witnesses and the stakeholders.

Benefits of DPoS include energy savings, greater decentralization, and positive participant behaviors. DPoS promotes greater democratization: all token holders can play some role in the operationalization of a network. It should be noted that in many models of blockchains that use DPoS, voting is proportional to the amount of stake that participants place. DPoS also aims to promote positive behaviors from participants. Those witnesses that add value to a network and are good network citizens are more likely to be repeatedly voted in; any witness that loses credibility, doesn’t participate, or looks to carry out fraudulent activity is likely to be voted out. This process ensures that witnesses work with the network’s best interests in mind. DPoS also makes use of “delegates” who are voted on to govern the system and propose any core changes. Like witnesses, delegates are placed into position through voting—and they can be removed through voting if they are seen not to serve the needs of the network.

In terms of financial services, and payments in particular, transaction rates are likely to be quicker in DPoS than in both PoW and PoS due to network engagement. This is because witnesses are likely to participate more actively and at greater speed to retain their position in the community. Witnesses are also likely to be limited in number, which increases transaction throughput; however, if there are an excessive number of validators, there is a greater risk that the network will slow down. So, a balance around greater scalability must be struck.

Although DPoS promotes greater decentralization, in some instances, such as where participation in the voting process is low, concerns about centralization can arise. For example, having a limited number of token holders could impact and influence the network. As with PoS, DPoS stakeholders can use their power from larger stakes to form cartels, which can make the network more centralized and less resilient to attacks—resulting in both regulatory authorities without relevant powers or tools to rectify such issues and concerns about market integrity and operational resilience.

DPoS is relatively new, yet it provides opportunities for firms to generate efficiencies through a relatively fast transaction throughput and can be delivered in both permissioned and permissionless networks. If developed and implemented in a compliant manner, DPoS can create positive outcomes for markets and consumers by providing interesting use cases in regulated activities, like payments. This is because it can create a truly decentralized environment with the potential for quicker transaction rates. That said, because the mechanism has not been tested as long as PoW or PoS have been, regulators should consider associated network-security risks, cartel-like behavior, and limited voter participation.

Practical and Istanbul Byzantine Fault Tolerance (pBFT/iBFT)²⁰

The consensus mechanisms explored so far work well when network participants are not trusted—that is, they work well with public or permissionless networks. This next section deals with those consensus mechanisms in which network participants are trusted; pBFT is among the most well-known examples. pBFT works best in cases where network participants are at least partially trusted—that is, in permissioned networks. It predates blockchains in terms of the underlying theory, but it has been applied to blockchain networks with some success. It can be considered a Proof-of-Authority (PoA) type of mechanism where nodes stake their identity and reputation, as opposed to a financial or computational stake such as PoW and PoS. This means identities are known, and therefore networks are likely to be centralized.

Nodes in pBFT are sequentially ordered and constantly communicating to keep network throughput high. There is one main node, also known as the “leader” node, and several backup nodes. pBFT assumes that some of these nodes are likely to be fraudulent and therefore likely to be sending out malicious or false messages. To counter this act, all nodes communicate with each other. The aim is that all the honest nodes come to an agreement on the state of the network through a majority consensus. As more honest nodes are likely to come to agreement on the state of a network than fraudulent nodes coming to agreement on a false decision, incorrect information should be rejected by the majority. Nodes in the system share messages when determining whether to commit a block to the chain. This ensures that a message came from a specific peer node and that it wasn’t tampered with during transmission. These messages take the form of four rounds, known as “views,” that are proposed by the leader.²¹ If a view has gone on too long, nodes can agree on a timeout. Leader nodes can be changed after every view.

Although pBFT is a pre-blockchain proposal for distributed systems, slightly modified for blockchain, iBFT was developed specifically for blockchain. iBFT can be thought of as a type of pBFT consensus mechanism with some modifications. Rather than using leaders and backup nodes, iBFT employs “proposers” who play a role like leaders and validators who play roles like backup nodes, in that they can validate blocks but play no role in the consensus protocol. After each consensus round, the validators can choose a new proposer who is responsible for providing the candidate block for the next interval. The primary modification in iBFT is that validators can be removed or added; in pBFT the set of validators is static. So, validators in iBFT are more invested in the network and likely more honest and engaged. iBFT may create lower barriers to entry depending on how the blockchain is implemented.

Both iBFT and pBFT work where fraudulent or malicious nodes do not exceed a third of the overall nodes in a network. This means that network security increases when the overall number of nodes in the network increases. The main advantage of pBFT and iBFT when deployed in financial services is that there are no forks and there is immediate settlement and so settlement finality. Unlike in PoW, PoS, and DPoS, which require confirmation of a block state, the stream of communication between nodes on a pBFT and iBFT network ensures that there is agreement on the state of a network at a specific time, which means the state of an agreed-upon block is final.

There is little energy consumption when compared to PoW-based networks, as there are no miners working to solve complicated mathematical puzzles and nodes are known to each other. Both pBFT and iBFT can lower the need for and variance of rewards, as decisions are made collectively; every node can be incentivized and rewarded in a similar fashion. Even though more nodes mean more network security, pBFT and iBFT work best when node numbers are limited because of the large number of messages exchanged between the nodes. The more nodes there are, the longer it takes for actions (like transactions) to be completed. Unlike the consensus mechanisms mentioned earlier, pBFT and iBFT work where nodes are known to one another; however, even with known nodes, a pBFT- or iBFT- based network can be susceptible to Sybil attacks in which a single node, like the leader, can manipulate other nodes in the network, therefore compromising security. While this risk can be mitigated with a network of more nodes, such expansion may reduce network speed and efficiency.

pBFT and iBFT can ensure efficient and fast networks and may protect consumers; however, the mechanism may facilitate networks that hamper competition. Within financial services, pBFT and iBFT are likely to work best where organizations represent the different nodes in a network and these organizations work within a single governance system. These organizations may create barriers to entry for new joiners, particularly where there is a closed loop of data sharing, sharing of efficiencies, or a closed ecosystem. This case is completely different from that of PoW, PoS, and DPoS, where almost anyone can become a node within a network, providing open, free, and contestable markets. It also means pBFT requires trust in some central entity or group of entities. This trust in intermediaries runs counter to a core tenet of the original Bitcoin Blockchain and most blockchains developed since. Decentralization and democratizing of actions (like currency transactions) are a key reason for the development of blockchains. Authorities may need to work with broader domestic regulators and competition authorities to mitigate against possible risks to competition.

Although most regulatory and technical risks are like those of pBFT, a network using iBFT will always produce blocks at regular intervals. This is true even if there are no actions or transactions. So, there might be many blocks with zero actions. Authorities may consider the implications of this type of block creation, as it can take up valuable storage space and create unnecessary energy expenditure. However, with “known” participants, authorities might be better able to regulate and supervise network participants, and, with immediate settlement, propositions built on such networks are more likely to ft into existing regulatory frameworks.

Federated Byzantine Fault Tolerance (fBFT)²²

fBFT is a Byzantine Fault Tolerant consensus mechanism that aims to solve issues of centralization with greater scalability. In the BFT mechanisms described above, identities of all participants are known; however, in fBFT the identities of all nodes do not have to be known. So, membership is open and control can be decentralized. fBFT can be thought of as semi-trusted as it relies on “quorum slices,” or a Unique Node List, which means only a subset of nodes is needed for agreement. In most implementations of fBFT, nodes individually choose the other nodes they trust for information. These nodes are likely to overlap or intersect with other quorum slices. Where this occurs it is known as a “quorum intersection.” Quorum intersections allow small quorum slices to build agreements across the network, improving scalability.²³

Within financial services, BFT consensus mechanisms can generate large efficiencies for retail payments, wholesale settlements, and other back-office functions, but BFT works best where network participants are known. These algorithms are built for speed and have immediate block finality. The latter is particularly operationally flexible, as it allows groups of validators to be modified over time. With increasing scale, speed and trust can become compromised; therefore, these models do have the potential to create barriers to entry. This problem is something that fBFT aims to solve with its approach of a Unique Node List.

Authorities might consider how fBFT consensus mechanisms can balance a decentralized and distributed network with efficiencies, such as a high transaction rate and settlement finality, to generate positive outcomes for markets. With its unique ability to mix known and unknown participants, fBFT could create greater risks to financial integrity than the BFT mechanisms described earlier, and it suffers from additional flaws and weaknesses that must be considered (most notably, in security, with the presence of faulty or malicious nodes). However, given its potential to be both efficient and scalable, authorities may need to collaborate both domestically and across borders to ensure that any risks to financial stability are appropriately mitigated.

DiemBFT²⁴

DiemBFT is a Byzantine Fault Tolerant protocol based on the HotStuf protocol, which itself builds on pBFT and aims to increase speed and efficiency. The aim of the protocol is to provide a quicker network by reducing the large number of messages and communication between nodes that are seen in pBFT, while retaining network security and accuracy. The HotStuf protocol achieves this goal through a type of communication that relies more heavily on the leader node, rather than on communication between all nodes, creating a star communication network. Security is also enhanced using the HotStuf approach of unpredictable leader election whereby a new leader is determined randomly. Like pBFT, HotStuf and DiemBFT are based on a leader/follower paradigm. In the DiemBFT Diem Network, members act as nodes and receive transactions from clients, which are communicated with others through a shared mempool.²⁵ Nodes can rotate and become leaders through a dynamic leadership model that is based on the HotStuf protocol, which allows these rotations to occur within rounds (or “views”) with minimum timeouts, and DiemBFT builds on this effort with a pacemaker mechanism to ensure every round has an upper time limit. Leaders propose new blocks of transactions to add to the chain and follower nodes vote to approve them. Once a node block gets a majority vote from the follower nodes, the leader node creates a Quorum Certifcate that is broadcast to all nodes. Once the Quorum Certifcate is verified, transactions can be committed to storage.

DiemBFT allows for faster transaction throughput. Transactions on the Diem Network have the potential to be relatively quick and low cost. Unlike the Bitcoin Blockchain, which can only process 7 transactions per second, the DiemBFT proposes to process 1,000 at launch. DiemBFT is also planning to support additional resources, such as decentralized applications, rather than just transferring Diem stablecoins.

However, the DiemBFT has several drawbacks that can impact competition as well as create issues around data and privacy. When used as part of the Diem stablecoin proposition, DiemBFT is a centralized network where nodes must commit at least $10 million to join and have access to appropriate computing hardware. And while these sunk costs reduce the incentive for those nodes to then become malicious, they create concerns around barriers to entry and market contestability, which run counter to the BFA’s aims. DiemBFT also relies on only a few counterparties (the nodes) as well as the leader node in each “view.”

Given the development of the DiemBFT by a “BigTech,” authorities have several unique and important challenges to consider. In addition to the high barriers to entry, the potential of the DiemBFT, when utilized by Diem Network members, to become a systemic payment infrastructure due to a potentially large embedded userbase (that is, service users of existing network members) may require extensive collaboration among regulatory authorities worldwide, including both financial services regulators to manage payments related risks, and non-financial regulators including competition authorities for broader risks (Bains, Sugimoto, and Wilson 2022; IMF 2021c).

Proof-of-Elapsed-Time (PoET)²⁶

PoET is a privately developed consensus mechanism that aims to prevent large energy consumption and can be used in both permissioned and permissionless blockchains. It also seeks to limit other forms of resource utilization—for example, locking up amounts of crypto assets as stakes and centralizing rewards. Like fBFT, it has been developed for a permissioned setting but has the potential to be scalable and extend to permissionless networks. Intel created this consensus mechanism and offers a ready-made tool to solve the computing problem of randomly selecting a leader. PoET makes decisions about mining rights and block winners on a network; it aims to spread the chances of winning fairly across the largest number of network participants.

Each node on a PoET network generates a random wait time and is required to wait for their chosen wait time to expire, which can save energy. The first node to complete the wait time—that is, the node with the shortest randomly generated wait time—wins the new block. Importantly, while the node is waiting for its wait time to expire, it goes to sleep, so it can use its processor for other tasks or no tasks at all, therefore limiting energy consumption. Intel’s ready-made tool ensures that networks using its mechanism can both generate random wait times and ensure the random wait time is genuinely fulfilled. PoET solves the problems with energy consumption, the potential centralization of rewards, and concerns about random leader selection. Furthermore, it can keep transaction rates fairly high.

That said, settlement remains probabilistic. Issues can arise with existing regulatory regimes when a PoET-based blockchain network is used in certain financial services capacities. Probabilistic settlement raises concerns about the ability of such blockchains to meet existing regulatory requirements on settlement finality as well as their use in certain financial service propositions (for example, payments). There are concerns about the network’s vulnerability to Sybil attacks, and giving centralization of the consensus mechanism to a third party (that is, Intel) runs counter to the goal of removing trust in intermediaries. Authorities should consider whether security concerns are likely to impact the provision of financial products and services based on PoET; they should develop relevant systems and controls for greater operational and cyber resilience. Although a single entity (such as Intel) is theoretically easier to regulate than unknown or distributed entities, regulatory authorities may find it difficult to draw an effective financial regulatory perimeter.