Hashing It Out On ICP: Proof-of-Work to Proof-of-Stake to Proof-of-Useful-Work

Published in

The Internet Computer Review

22 min readDec 6, 2023

“If you look at a thing nine hundred and ninety-nine times, you are perfectly safe; if you look at it the thousandth time, you are in frightful danger of seeing it for the first time.”
― G.K. Chesterton

I- How To Hash Ourselves Out

Regardless of your blockchain netizenship, we all enjoy watching a masterful culinary performance.¹ Part of the joy is watching a great chef composing ingredients into a delicious meal; ingredients that may oppose each other in flavor but together maximize each other’s pleasurable tastes and negate each other’s displeasurable aspects. Like a musical composer, high notes lead to low notes and a symphony for the taste buds is formed. It is an enrichment.

To a blockchain network aficionado, hash is a particularly interesting dish to envision. You can make hash out of almost anything–with no recipe. That is the point of it. Arguing about what is the best ingredient for hash would be silly indeed. You use what you have. It may be, given the range of ingredients you have–if you have enough competency–you could navigate the composition of a meal well enough to make a delicious one. You could just play it safe, do it the way most think is the best way. But, there is a willingness to venture to find the truth. To hash over² a problem is to discuss thoroughly, bringing in different points of view and possibilities of a solution and find–not just a solution–but a meta solution that transcends the previous objectives.

Although Vitalik Buterin, the inventor of Ethereum, was a founder of a major Bitcoin publication and Dominic Williams, the CEO of Dfinity, initiated his researches in the Ethereum community, the thought of the Bitcoin and Ethereum and ICP communities treating each other in an accepting let’s-all-have-a-meal-together manner today seems too far fetched for reality.³⁴⁵⁶⁷⁸⁹

So it seems, we should all just hash it out. What are our ingredients? Let’s take a look. The fate of the metaverse depends on us.

What are we trying to hash out exactly? We are trying to hash out how people will agree in the future. A worthwhile discussion. A smart contract¹⁰ on an open distributed computing system would offer a higher rate of determinism for agreements than a traditional contract using a pen-and-paper-and-promises system. So, how should we go about creating a suitable distributed computing system for all to use?

To clear things up, we must address that in a distributed computing system–like a blockchain–there are schemes that function as a consensus mechanism (a way with which computers in a network agree) and schemes that function as a sybil resistance mechanism (a way with which membership of computers in the network is limited). To reliably function, a consensus mechanism relies on the sybil resistance mechanism to limit the number of participants that are allowed into the system.

``A distributed system is one in which the failure of a computer you didn’t even know existed can render your own computer unusable.’’ –Leslie Lamport¹¹

A consensus mechanism¹² helps individual participants come to an agreement on a function, like the order of transactions initiated by participants. If you change up the order of the transactions without coming to a settled agreement, an attack called a double-spend attack¹³ happens. The same token sent to a particular address is sent again to a different address. This breaks the functionality of a shared transaction ledger.

All consensus protocols have fault bounds. They are guaranteed to work in the presence of faults that could be participants in the network acting as offline nodes or malicious nodes. So long as a consensus protocol is within its fault bounds, it will function as intended. Should the number of faulty nodes exceed the fault bounds of the protocol, then an adversary might then cause the protocol to cease its intended functionality.

“3 E’s Sybil Resistance 1. Entry Cost 2. Existence Cost 3. Exit Penalty.” –Dominic Williams¹⁴

A sybil resistance¹⁵ mechanism is needed to prevent an adversary from creating multiple faulty nodes to overcome the fault bound. The name for this attack was inspired by a novel written by Flora Rheta Schrieber about a woman with dissociative identity disorder,¹⁶ a mental illness where multiple distinct identities in a person’s psyche disrupt their cognitive functions. Much like how multiple instances of nodes that are faulty would disrupt a distributed systems’ consensus functions.

A structured means of analyzing the effective dimensions of a sybil resistance mechanism is proposed by Dominic Williams in the concept of the 3 E’s of Sybil Resistance:¹⁷ Entry cost, Existence cost, and Exit penalty.

Entry cost is the requirement that a cost of resources be incurred in order to initially participate in a consensus protocol. Other than a monetary cost, time and effort is also noted to be part of the cost of resources.

Existence cost is the requirement that a cost of resources be incurred in order to maintain participation on a continuous basis in a consensus protocol. This cost requirement prevents participants from continually investing into new identities and provides a more decentralized influence over the network as participants are bounded in the number of identities they can afford to support.

Exit penalty is the consequence incurred for faulty behavior by a network participant that results in expulsion from the network. This requirement faces an adversary with a possible unbounded cost of attack. An adversary should not be able to recycle the resource cost of attack into another instance and renew the attack ad infinitum.

Now that we have a framework for understanding how agreement between computers form, we can now ask how the Bitcoin and Ethereum and Internet Computer Protocol networks hash themselves out.

II- Bitcoin hashes itself out using Proof-of-Work

“Proof-of-work has the nice property that it can be relayed through untrusted middlemen. We don’t have to worry about a chain of custody of communication. It doesn’t matter who tells you a longest chain, the proof-of-work speaks for itself”
-Satoshi Nakamoto

Bitcoin¹⁸ is a simple protocol, optimized to do only one thing: to act as an append-only transaction ledger to transfer value over the internet.¹⁹ Its very invention is a reaction to the shifts in monetary policy imposed by governments and executed by banks.²⁰ The creation of Bitcoin by Satoshi Nakamoto presented a crypto-economic consensus mechanism and a sybil resistance mechanism²¹ at the same time, which is called proof-of-work.

At its heart proof-of-work is a competition. In order to append a new block of transactions to the ledger, a miner must find the value of a “Nonce” that will result in a hash of the block starting with some number of leading zeroes.

A block in the Bitcoin protocol is a data structure that contains a number of data fields that a miner may not alter except for a value called a “Nonce”. Short for: number only used once, a miner alters the nonce to generate a hash using the SHA-256 algorithm until a hash that meets the difficulty target is generated. Network difficulty of Bitcoin is adjusted according to hash power by adding or subtracting a leading number of zeroes to the target hash in order to decrease or increase the probability of a miner successfully meeting the difficulty requirements.

The hash generated by the SHA-256²² algorithm is a value that is random and unpredictable. There is no shortcut to finding a nonce that fits the target hash. A participating miner must spin the nonce, through expending energy, trying different values until a matching hash is discovered and the winner is awarded the ability to add a block to the Bitcoin ledger, including the block reward.

Sybil resistance is occurring because in order to find a matching hash you must expend computation power–through electricity–in order to puzzle out combinations that enable you to append a block and receive the block reward.

Because the relative proportion of the block rewards a miner receives is proportional to the number of hashes the miner produces, the game is to perform this hashing out as efficiently as possible. For hardware, you need an ASIC (Application Specific Integrated Circuit) that is specialized for mining for hashes and for energy you need a cost-efficient source of electricity to power the hardware. Sybil resistance entry cost would then be the purchasing of these ASIC hardware miners and the existence cost would be the enormous cost of electricity it takes to power the ASIC hardware. Exit penalty would be the loss of premium should the mining ASIC rigs be sold, which is a relatively weak penalty.

In theory,²³ an adversary that would break the Bitcoin network by overcoming the fault bounds would then need to secure ASIC hardware and a cost efficient energy source to generate up to 51% of the current hashing power generated by the Bitcoin network. A spawn camp attack²⁴ would then be possible to render a proof-of-work blockchain useless.

When you take a hash²⁵ from SHA-256 of some input bytes, the output is random and cannot be predicted. Therefore the hash is a random number generator. The realization is that stateful decentralized networks are driven by random numbers.²⁶ Satoshi’s innovation of public crypto economic consensus²⁷ leads to other forms of securing decentralized networks.

III- Ethereum hashes itself out using Proof-of-Stake

“The ‘one-sentence philosophy’ of proof-of-stake is thus not “security comes from burning energy”, but rather “security comes from putting up economic value-at-loss”.”
-Vitalik Buterin

Ethereum²⁸ is a platform built to host a new type of software called a “smart contract”. Similar to a vending machine,²⁹ a smart contract is defined by a predetermined input that results in a predetermined output, executed in an autonomous way. Although Bitcoin is arguably a protocol smart contract, it is the Ethereum protocol that offers a turing complete programmable smart contract using a language known as “Solidity” as a more flexible means of defining particular inputs and outputs of a digital agreement.

The Ethereum Virtual Machine³⁰ (EVM) is the environment where smart contracts live. The Ethereum protocol keeps only one canonical state that is shared by all the nodes in the network. If Bitcoin acts as a transaction ledger, Ethereum acts as a state machine to the internet. The “turing complete”³¹ aspect of this state machine means that it can be used to solve any computational problem given enough time and resources.

Decentralized finance³² (DeFi) applications that enable exchanges, lending, stablecoins, derivatives, and other uses become possible as well as other applications such as NFTs³³ and DAOs³⁴.

Ethereum has the potential to act as a global settlement layer for DeFi transactions.³⁵

Originally using proof-of-work as its consensus algorithm, Ethereum switched to proof-of-stake during “The Merge”³⁶ after many years of research and development due to the cost of expense to maintain and secure the protocol in its original configuration.³⁷ Consensus mechanism and sybil resistance mechanism became less coupled together allowing for separate optimizations on both.

The consensus mechanism for the present Ethereum is named “Gasper”³⁸ which is a combination of Casper FFG³⁹ and LMD-GHOST⁴⁰ rules for consensus. LMD-GHOST is used to determine the legitimate state of the blockchain by confirming the blocks that have the most calculation and latest messages. Casper FFG is used to finalize the confirmed blocks to result in a tamper proof permanence to the blockchain consensus.

The proof-of-stake sybil resistance mechanism is a scheme where network nodes, called validators, participate in sybil resistance by staking a predetermined amount of tokens (32 Ether) then, provided the nodes follow the consensus rules, are rewarded with tokens in proportion to the amount that was locked-up. Nodes that violate⁴¹ the consensus rules are “slashed” causing them to lose their staked tokens.

A “validator” is a virtual entity in proof-of-stake that replaces the “miner” in proof-of-work. A validator is responsible for storing data, processing transactions, and adding blocks to the Ethereum blockchain. In the absence of the random hashes generated by proof-of-work, validators in proof-of-stake need to rely on an algorithm called the RANDAO⁴² (Random Decentralized Autonomous Organization) to pseudo-randomly select a validator in a fair and verifiable way to propose a block to be appended to the blockchain.

Sybil resistance entry cost for Ethereum proof-of-stake would be the capital needed to purchase the ether needed to run a validator node. Existence cost would be the cost of bandwidth needed to maintain the instance of a validator node, which is a relatively weak cost. Exit penalty has circumstances defined in both the LMD-GHOST and Casper FFG aspects of consensus. “Slashing” occurs when a validator violates specific rules in the Ethereum protocol that could be part of an attack on the chain, or “Byzantine behavior.” Slashing conditions⁴³ relating to Casper FFG impose a cost of attack on reverting finality of the blocks and slashing conditions relating to LMD-GHOST impose a cost on the “Nothing-at-Stake attack”⁴⁴and “Balancing attack.”⁴⁵ On both types of slashing conditions a byzantine attacker is imposed with an unbounded cost of attack.

Proof-of-stake, despite its advantages relative to proof-of-work, has some inherent challenges.⁴⁶ Because validators are virtual instances without need for dedicated ASIC hardware, a network node can be created in a cost-efficient⁴⁷ manner on a corporate cloud service like Amazon Web Services⁴⁸ where there is potential for the cloud service provider to interfere with a validator or close them down. Another is that a “short selling attack”⁴⁹ on a proof-of-stake system may be possible using traditional short selling in tangent with an acquisition of a large percentage of ether tokens. A sybil attack against proof-of-stake would also happen instantaneously since an adversary would not have to spend time configuring any hardware to initiate the attack.

Between proof-of-work and proof-of-stake it is debatable⁵⁰ which sybil resistance mechanism is superior, but might it be preferable to have both as they are strong against different types of attacks?⁵¹ If so, other considerations may be presented.

IV- ICP hashes itself out using Proof-of-Useful-Work

“Proof-of-useful-work creates sovereign blockchain networks that are hosted by dedicated hardware. In combination with deterministic decentralization, and special cryptography, it can create horizontally-scalable secure public blockchains with extraordinary efficiency”
— Dominic Williams

The Internet Computer Protocol⁵² (ICP) is meant to act as a world computer⁵³ for the internet. It aims to replace the traditional IT stack–where a developer can fall prey to the dangers of platform risk present on the corporate cloud infrastructures⁵⁴— with its decentralized cloud infrastructure. The ICP works by utilizing dedicated nodes run by independent data centers that use the ICP to message and collaborate to form a virtual global computer. The nodes are governed in subnets by an advanced DAO called the Network Nervous System (NNS). The subnets then host bundles of smart contract code called “Canisters”⁵⁵ that are the building blocks of dApps hosted on the ICP. Code in canisters are compiled to Wasm⁵⁶ from a variety of possible languages including ICP’s own native language, “Motoko.”⁵⁷ To orchestrate the execution of canister code and the nodes that authenticate the network state, a set of cryptographic protocols are used that are called “Chain Key Cryptography.”⁵⁸

Dominic Williams sought⁵⁹ to create faster blockchains by repurposing traditional Byzantine Fault Tolerant (BFT) consensus mechanisms⁶⁰ for a decentralized setting.

Consensus algorithms⁶¹ have the properties of safety and liveness to measure their efficacy. Safety means that nothing bad happens. Liveness means something good will eventually happen.⁶² Traditional BFT protocols rely on rounds of message passing that provides instant safety over liveness.⁶³

There are three different types of Byzantine Fault Tolerant protocols:

Synchronous BFT protocols assume that all messages are delivered in some amount of time and that all participants know how long it would take for the messages to be delivered. When network asynchrony is sufficiently high like during a Ddos attack,⁶⁴ synchronous protocols become unsafe and fail to output a value. While it was never directly characterized in the whitepaper, the Bitcoin nakamoto consensus can be placed in this category⁶⁵ ⁶⁶.

Partially Synchronous BFT protocols assume that all messages are delivered in some amount of time and that all participants do not know how long it would take for the messages to be delivered. When network asynchrony is high, partially synchronous protocols remain safe but also fail to output a value. The Gasper consensus mechanism of Ethereum is unique⁶⁷ as it contains aspects of a synchronous protocol in the LMD-GHOST which sets strict timing conditions⁶⁸ and aspects of a partially synchronous protocol in Casper FFG which relies on epochs⁶⁹ to decide on finalization.

Asynchronous BFT protocols do not make any timing assumptions on network messages and thus are not susceptible to Ddos flatline attacks. Asynchronous protocols rely on a random number generator to direct message passing between participants. Dominic Williams concluded that eventually-consistent, blockchain-like protocols driven by random numbers generated by cryptography were superior in a decentralized setting.

The Dfinity Foundation, led by Dominic Williams, started⁷⁰ to generate random numbers using “Threshold Relay” to generate randomness which uses BLS Signatures to aggregate⁷¹ a source of randomness⁷² that is incorruptible, unmanipulable, and unpredictable without using a consensus mechanism. This Random Beacon Chain is then used as a stepping stone to reach other forms of consensus.⁷³

The sybil resistance mechanism of the ICP is called proof-of-useful-work. Rather than the competition that is proof-of-work, proof-of-useful-work is a cooperation.

ICP is created by hardware nodes like ASIC miners in Bitcoin, and unlike validators in Ethereum which can be created on virtual machines using the corporate cloud infrastructure. These are dedicated node machines⁷⁴ that process transactions and calculations.

These node machines are grouped together in what are called subnets and are incentivized to stay within a productivity bound to produce the same number of blocks as the other nodes in the subnet. The random beacon chain is utilized to select rank individual nodes to act as a leader and produce a block⁷⁵ by a mechanism called Probabilistic Slot Protocol⁷⁶ (PSP). If that selected node fails to produce a block then the node next in rank gets a chance to produce a block. A node that is built to the same standard specifications as the other nodes in the subnet should be able to produce a block when it is selected by the random beacon. Because the random beacon decides which node gets to produce a block, a node built of higher specifications than the rest of the subnet would not have any advantages in getting a higher level of reward than the other nodes. A node built of lower specifications may not be able to produce a block when the network is under load and would get a lower level of reward than the other nodes or may even be slashed and removed. In this way the ICP network prevents nodes from statistically deviating. The chain key cryptography protocols enable ICP to add and remove nodes from the network in a graceful manner without interrupting its operation.

The Network Nervous System (NNS) is a key piece of infrastructure in ICP that acts as a governance mechanism⁷⁷ for the protocol. Rather than rely on social coordination by a foundation or company to organize protocol management you have a DAO designed for on-chain governance where voting is recorded on the blockchain ledger and voting rules are transparent–being pre-defined and communicated in advance. Protocol updates by an on-chain DAO are embedded in system code and would be implemented immediately upon ratification. NNS DAO Governance can take special cases into account and prevent nodes from being slashed for deviating for bonafide reasons that may happen through no fault of the node operators⁷⁸ ⁷⁹.

Voting in the NNS involves users locking their tokens into what are called “Neurons”⁸⁰ which enables delegated voting using a follow system and a flexible weighted voting system where a voter can choose how long to lock up their tokens and the longer they choose to, the more voting power is attributed to their neuron. Staking rewards are then distributed to voting participants to incentivize value participation. It’s important to note that although the NNS governance system might look similar to a proof-of-stake system, that it is a governance mechanism and not a sybil resistance mechanism for consensus. Although the NNS can be called the most advanced DAO in the world, important upgrades can still be made in the form of the use of Trusted Execution Environments⁸¹ for node hardware to improve network security⁸² and the use of People Parties⁸³ as proof of personhood to improve the voting mechanism to favor verified humans.

The sybil resistance mechanism of ICP proof-of-useful-work depends on the random beacon chain, probabilistic slot protocol, and the NNS. All 3 E’s of sybil resistance are achieved by proof-of-useful-work. Entry cost is the required acquisition of hardware node machines that must be added through governance of the NNS. Existence cost is the hosting and operation of the hardware node machines so that they function within the productivity bound of their subnet. Exit penalty for the ICP involves slashing directed by the NNS of the faulty nodes that provides more circumstantial precision.

It is the intention of ICP to integrate both with Bitcoin and Ethereum to both benefit its own systems and improve the systems of its predecessors. Making use of novel threshold cryptography⁸⁴ designed⁸⁵ by the team of cryptographers working in the Dfinity Foundation, canisters in ICP can sign transactions on the Bitcoin and Ethereum blockchain without the use of a central custodian (like a bridge) that may be exploited.⁸⁶

Integration with Bitcoin has already happened.⁸⁷ Bitcoin integration on ICP enables smart contract functionality transplanted on the Bitcoin network, enabling a DAO⁸⁸ governed DEX⁸⁹ that can manage Bitcoin and other use cases like SocialFi⁹⁰ and GameFi⁹¹ with Bitcoin. ICP gets to benefit from the large liquidity that is present in Bitcoin.

Ethereum could be supercharged by integrating with ICP.⁹² The scalability and security and efficiency of interacting with the EVM would be improved by offloading compute intensive processes to ICP canisters and then having the settlement logic execute on the Ethereum base chain similar to the way L2 rollups⁹³ work. The reverse gas model⁹⁴ on ICP would mean that the user interface for Ethereum dApps would be made more efficient in that users will not have to pay upfront just to interact with the Ethereum smart contract. Defi applications on Ethereum are secure on the back end but for users to interact with them they must pass through vulnerable front ends built on corporate cloud providers.⁹⁵ ICP and Ethereum integration would make it possible for an Ethereum dApp to use ICP sign in with Internet Identity⁹⁶ to authenticate a user more securely. ICP canisters would be able to serve the Ethereum dApp website directly to users without the use of a corporate cloud platform.⁹⁷ Integration would also enable the formation of fully autonomous DAOs in Ethereum using the SNS⁹⁸ infrastructure. As features evolve on ICP, eventually, the hope would be to allow Ethereum validators to host their nodes on ICP to solve the platform risk issues currently present for proof-of-stake participants.⁹⁹

With such tooling made available it will be up to the courage of the community to understand the underlying protocols and build systems that will support digital sovereignty that will be much needed in the coming era.

V- Hashing the metaverse out

“This is not a simple space. But the fact is, just like Ethereum added value to Bitcoin, the Internet Computer can add huge value to Ethereum, which I believe is here to stay, and vice versa. Ethereum is crucial to the entire ecosystem, including the Internet Computer. This is not a zero sum game.”
– Dominic Williams

The act of eating together is a civilization test. Imagine all the societal agreements that must succeed for you and your societal network to have a shared feast. A feast is an act of sustainment and reciprocal altruism. Because it is not guaranteed that a single person, no matter how capable, will always catch enough food to sustain themselves, it is prudent while you can, to store the spoils of the day’s hunt in your friends by initiating gratitude through the act of sharing. In this sharing of abundance you enable a more robust protocol of sustainability that is of an advantage to you and an advantage to others. A win and help win scenario.

Like the creation of a meal of hash, each well chosen ingredient added makes the meal more likely to enable sustainment, a case can be made that having all three of these types of sybil resistance in networks is preferable since each type is more robust than the others against specific types of attacks thereby, the existence of each makes the others more robust.

A common argument among proponents against the concept of Bitcoin’s consensus protocol is that the demand of proof-of-work to consume energy is a waste of energy¹⁰⁰ and is therefore unsustainable. We have learned from developmental psychologists¹⁰¹ that iterated games that come to a consensus agreement are more sustainable than imposed power because imposition of power is a waste of energy. The imposition of my-blockchain-is-better-than-yours is the true waste of energy.

The word “hash” which is a key concept in these agreement protocols alliterates with the word “hack” and their concepts–in ways–taste similar too. To hack is to use a tool to its full potential by the will of intention. And a hacker is a person who enjoys exploring the details of programmable systems and how to stretch their capabilities, as opposed to most users, who prefer to learn only the minimum necessary. In hashing together Bitcoin and Ethereum and ICP, we can enjoy exploring the details to stretch their capabilities into a world network we can all use to attain our full potential in pursuing our personal challenges.

A blockchain network is an ideal tool to enable liberty and flee tyranny in that it has potential to solve fundamental organizational dilemmas in the human condition:

The king maker’s dilemma is the propensity of a trusted middle party to profit by acting as an intermediary to facilitate agreements. The more powerful the intermediary, the more leverage it has to extract value. Kings are favored because in disputes where there are no judicial intermediaries, there must be settlement in diplomacy or war. A blockchain network’s inherent characteristic of decentralization leading to determinism allows a network of intermediaries to disintermediate the model of a single trusted middle party.

The tower of babel dilemma is the propensity of an organizational hierarchy to become more opaque the more the hierarchy grows. A centralized leader on top of a tall tower is not likely to consider the lower rungs of a social organization. Blockchain’s inherent transparency would mitigate confusion caused by disjointed information as a social organization grows.

Return to violence dilemma is the propensity of a party with superior weapons to use violence to extract value from a party with inferior defenses.The cryptography that underlies control of digital assets on a blockchain means that ownership can now be turned into pure information. Cryptography gives the advantage to the defender and imposes a cost on the rubber hose attack¹⁰² adding friction to the scaling of violent attacks.

As in technological cycles past, we will witness a dramatic unbundling and re-bundling of how society functions such that it affects how people think and the very quality of their lives. We would do well to understand that negotiation against a tyrant is only possible with enabled sovereignty¹⁰³ and to enable sovereignty it is imperative to increase the base trust layer of society so that people can collaborate at scale. The check on power of corporations and governments is property rights and decentralization. Networks will be as meaningful as religions and states.¹⁰⁴

With the truth of this in mind, we can look forward to a future where we can all sit side by side at a table, carefully discussing the beautiful future of the metaverse, over a feast of hash everyone helped make together.