Measuring Decentralization: The Nakamoto Coefficient Explained
Definition of Nakomoto Coefficient
Almost all blockchains claim themselves to be decentralized but how decentralized their blockchains are is not something an average crypto layperson understands. Luckily, there is a way to measure one of the most important aspects of decentralization and assign them a score, the methodology of which we will discuss here. One of these is the Nakomoto Coefficient (NC) and Solana was one of the first chains to popularize this term. In this article, we will discuss decentralization largely in the context of nodes in Proof of Stake (PoS) networks and sometimes contrast it with Bitcoin Proof of Work (PoW) mining. The Nakamoto Coefficient is commonly known as the Minimum Attack Vector (MAV) on PoW networks.
The Nakomoto Coefficient is essentially a measure of the number of nodes it takes to break a typical Proof of Stake (PoS) consensus network that can lead blockchains vulnerable to 51% attacks. A compromise to a network is possible when anyone with malicious intent attaches 51% of nodes and is able to asset power unilaterally. Given the current state of blockchains, it is likely that if more and more node operators gather together and decide to break the integrity of a network, they can do so by replacing information on the blockchain and reversing the transactions and rules that dictate how a network should be maintained. Basically, an NC score is assigned for most of the blockchains that are maintained by nodes. The higher the NC score in a relational sense and the better decentralized a network is considered as per the industry standard.
For some blockchains the minimum threshold to break the network can even be 70% to 80% in some cases in order for a bad actor to take majority control of the blockchain. So it’s not always 51% but that is the baseline standard we work with to calculate the Nakomoto Coefficient. How the Nakomoto Coefficient is calculated depends on six different levers and the following are brief explanations of those.
Mining (also known as staking in PoS networks): Participating in staking or mining rewards users with a % return on a monthly basis.
Validator client codebases: Most Bitcoin validators run on Bitcoin Core which basically tilts towards a high degree of decentralization.
Development: The number of developers who have made commits. In the case of above mentioned Bitcoin Core, there are relatively few engineers.
Exchange: The volume of trades that take place across exchanges within a set period of time usually 24 hour periods. (On a side note, from personal expeirecne we have noticed that some of the Cardano native DEXs and other EVM based ones are not accounted for on Coinmarketcap, so the volume figures there are not the most accurate but the most reliable.)
Nodes: The node distribution across geographies.
Ownership: This measures the distribution of coins with a minimum threshold value converted from a standard $ value and see how distributed those owners are across wallet addresses. This is interesting because it signs the possibility that the government will only need to confiscate wallet addresses associated with key individuals to get hold of their crypto assets which can get the prices tanking.
Source: https://news.earn.com/quantifying-decentralization-e39db233c28e
After plotting personal information as a Lorenz curve on a graph, you can determine the minimum number of entities needed to cause a significant disruption in the system. You need to find the smallest number of entities whose combined control equals or exceeds a specific percentage i.e. 51% for blockchain network disruption. The higher the resulting score in the Nakamoto Coefficient, the greater the level of decentralization in a blockchain.
History of the Nakomoto Coefficient
The metric was first put forth by the former CTO at Coinbase, Balaji S. Srinivasan, in a medium article titled Quantifying Decentralization. They named the coefficient as such in honor of the pseudonymous founder of Bitcoin, Satoshi Nakamoto. Minimum Attack Vector is a similar calculation concept focussed on Minining via a vis PoS networks.
The Edinburgh Decentralization Index (EDI) is an alternative method used for measuring the decentralization of cryptocurrencies. This method is currently employing Bitcoin and Ethereum chains for research and comparing how decentralized both are. Once the results are out, we will see the status of other chains. Not much is known about EDI currency to the public.
Ways to break blockchain networks
What makes a blockchain or blocks-producing chain is the number of nodes or validators in a given network. Attacking the majority of nodes can effectively dismantle a network. Below are the number of scenarios from the devil’s advocate angle that can threaten the integrity of a network:
Regulatory enforcement: A regulator can impose laws that can prevent operators from running nodes. There we have to consider the geography of the distribution of nodes as well. In China in the year 2021, the government banned mining facilities of Bitcoin fields. Node operators can still run a network in a banned country because eventually, it’s just a bunch of codes put together that go into running a PoS node. Nowadays, there are VPNs to dodge IP tracking. However, the people can still risk getting fined heavily and fear the consequences of violating the local laws.
In Solana’s case, according to validators app most of the validators are based in the US, so if the government decides to impose stringent regulations that could prevent the people from running nodes, it can quickly become a problem.
Internal or external attacks: Most node operators can conspire to break a network, hackers can attack nodes or viruses can infect the code that goes into maintaining a node because at the end of the day, there is someone in power as a centralized authority saying yes or no to what code goes into running a validator.
A large scale gang up of nodes to overturn transactions: In the year 2019, some 7000 Bitcoins were stolen by a hacker in a major hack on Binance. The CEO of Binance, CZ sought to roll back the transaction by suggesting re-org of Bitcoin nodes. The node runners were initially in agreement with the idea and this would have only taken 51% of them to agree to roll it back. However, this didn’t happen due to delays caused by the pushback from the industry players. And because of delay the blocks carrying records of that transaction were already produced. If you look at the possibility of doing it, if CZ acted rapidly to get that 51% the “unhack” could be perfectly possible.
“Not every Japanese can do Karate” There is a misconception that the more nodes there are the better. Not always! Not all nodes function and contribute equally. Defying the Nakomoto Coefficient, only a small percentage of nodes contribute to the consensus of a network. It is the nodes that produce most blocks that are key. In PoS networks, the nodes that produce most blocks performance wise are the ones where the holders of coins decide to stake and delegate their resources.
Centralized exchanges like FTX: If a holder doesn’t have their assets protected with hot wallets like Ledger and is instead saved in centralized exchanges, there is a chance that another FTX like case could happen whereas someone with bad intentions can play with people’s savings and steal it for use of illicit proposes. If an attacker has control over most of coins in circulation obtained by hacking a centralized exchange, they can effectively shut a network down. What happens is that if the coins are then held into nodes that the attacker can cut from the internet from those nodes, the entire network gets compromised.
Sybil attacks: In the context of PoS networks, a party can create several nodes under false identities to exert disproportionate control over the consensus of the network. With that they can overturn transactions, roll back on the already executed transactions, pervert data, spend a token or coin twice and play tricks on users. Requiring node operators to identify themselves using zero-knowledge identity proofs before they can create more nodes than should be allowed could be necessary in the future.
Mass hack of nodes: Say on Solana, we can consider the collective efforts of attackers required to gain control of 51% of SOL coins. It is important to note that this measure does not focus on the number of stake pools or nodes this time since a single party can manage multiple nodes. The key factor here is the stake held by the entity, not the number of pools responsible for block production. With that said to mitigate such a challenge, if stake pool operators run their pools on different infrastructures and have multiple validator client options to choose from it may get harder for attackers to get control of the majority of SOL and hence compromise the nodes simultaneously.
Whales: Large whales on PoS networks can throw curveballs making it all a little centralized too. Whales are parties that are responsible for ownership of large amounts of crypto assets. Whales can decide to run their own nodes and choose which pools they like to delegate their stakes to. Don’t be fooled by the distribution of stakes here, we are talking about how distributed the supply of collins can be. Whales who may seek to compromise with the integrity of the network and play the game in their favor if they choose to do so. As per the Supra Orcales report ADA is currently owned by more than 40% of all ADA holders and it’s the same with Solana because around 40% or so is held by insiders and the foundation. Therefore the distribution stake by country as claimed by the Solana Foundation to assess how decentralized they are in geography is something you should take with a grain of salt.
Source: https://solana.com/news/validator-health-report-march-2023#en
Frictions between blockchain scalability and decentralization
Since November 2023 post-FTX collapse, the concept of decentralization has changed so much, and the nitty gritty details matter more than ever. For a network like Solana that has adopted and popularized the concept of decentralization, there are deeper societal and technical aspects that need to be thought through as well. Not to mention if you are looking for a scalable blockchain, then decentralization is not that perfect (more on this later). Full on decentralization certainly would affect the network’s scalability as more validators are then in charge of validating the number of transactions which can strain the speed.
Less decentralization means high scalability
More decentralization creates friction with scalability therefore the trade off between the two is quite a fragile thing. Most people understand high scalability as something that is able to handle transactions at greater speeds and at the lowest possible costs. For decentralization we will need more and more nodes and the higher the number of nodes, the higher the costs to maintain those nodes. This implies that with greater maintenance costs come higher transaction fees in a network that create less of an incentive for operators to invest their time and money to run nodes. On the other hand, if the scalability is low and the maintenance costs are high, fewer people are willing to participate in a network meaning with fewer people more centralization comes about.
A network consensus can be censored by bad actors
Pragmatically speaking, decentralization is really important. Imagine a scenario where one node operator Bob decides to manipulate the codes of their Cardano stake pool by making some add ons not proposed by IOG engineers and cheating the system by censoring transactions and showing partiality towards certain exchanges. If only Bob does it, nothing happens but what if Bob, Alice, Jake, Jillian, Keith and many more come together to form a 70% majority and decide to change the original rules of the consensus protocol, then of course the rules preferred by the majority will get executed on the Cardano blockchain.
They could then ignore the blocks produced by the 30% minority and pass their own agenda, such as APY% payouts to users, censoring certain DEXs, or enforcing asset lockups. What that means is that the users who invested in ADA coins weren’t asked for their consensus. If the users are not interested in lockups and want liquid staking, they might lose faith in ADA as an asset and invest it elsewhere, or worse, leave the blockchain altogether. This scenario of course presumes that Bob and co. don’t hold as much ADA as the majority of users on Cardano. So basically when you give power back to the hands of users, the users get to decide whether they entertain a certain stake pool or not. If they are unhappy with a stake pool operator, they can go ahead and stake their assets elsewhere.
Pros of the Nakamoto Coefficient
Quickly come to pin point the blockchains and their decentralized levels by comparing their NC scores.
Know quickly which of the six levers that add to decentralization make more sense to you. For example if you want to become a node runner and the Validator Client metric is more important to you, you don’t have to choose a blockchain that leaves you with one client option.
You can identify the security threats to your blockchain such as legislation and see how where most of the node runners are spread out geographically. If the majority of your blockchain nodes are based in Australia, you can easily know why your NC is low and then do something about the problem.
Gives developers the ability to run several experiments on testnets before going live and check beforehand under which scenario the decentralization will make better sense.
Cons of the Nakomoto Coefficient
The coefficient calculations and graphs are way too complex and graphing them by applying a Lorenz curve is no easy feat. The issue is that there is not a standard formula that you can use to calculate the score so you need to collect the large mathematical data sets before you plot them on a chart.
Data set selection greatly influences Nakamoto scores and the selection can be manipulated. For instance, if you consider all small wallets, the blockchain may appear highly decentralized. Conversely, if you focus only on owners with large sums like $1000, the blockchain might seem very centralized. Validators can even create multiple wallets to appear decentralized.
Furthermore, the Nakamoto Coefficient, although valuable, falls short in terms of qualitative analysis which may not be easily quantifiable. As previously mentioned in this article, we have explored some of these nuanced thoughts in earlier sections. Additionally, the governance and decision-making aspects are further points of concern when it comes to the limitations of this key performance indicator, as discussed previously.
Most validators in the Cardano and Solana ecosystems are hosted on AWS and the likes of Hetzner. This is not new to just these two chains. This is what we mean by being wary of another single point of failure.
Source: https://solana.com/news/validator-health-report-march-2023#en
While the Nakomoto Coefficient is far from perfect and can’t truly represent the true nature of decentralization. It just takes a single point of failure for the entire network to break. With the current NC scores of BNB Chain, Near protocol, Cosmos and Celestia, they are no match to Solana or Cardano for that matter. It’s a good thing that more chains are starting to calculate their NC because it sets a benchmark and creates a competitive playfield for the chains to demonstrate their network distribution. Only with more think tank discussions and measuring formulas like NC can we as an industry compete for secure permissionless distributed networks.
However, despite the complex nature of the topic the first step forwards and the road to a more decentralized world of crypto, we need to be able to measure decentralization. Things are accomplishable step by step. As the saying goes “What gets measured gets managed” or in German "Alles mit maß und ziel” Overall the Nakomoto Coefficient is a good starting point for starters looking to understand what decentralization even means.
