The disagreements between the ‘big blockers’ and the ‘small blockers’ in Bitcoin are heating up. Bitcoin Classic is poised to release its first client to compete with Bitcoin Core, and Bitcoin Unlimited has had its first vote on its new feature set. It is a time of peril in the galaxy…
Now as the credits fade into the star field background picture a big wedge shaped Star Destroyer with the banner reading “Decentralization” filling the screen. This word is really the Battle Cry of most crypto-currencies, and as I have written in the past, it is so poorly understood.
It is a repurposed term, that simply describes a quality of network topology, transformed into a rallying call of rebellion. The problem is that almost everyone that I read or encounter in the industry uses this term as a panacea for all the problems that they see in the world today, without actually knowing what it truly means. They believe it because of faith from authority, and through basic reasoning, that it is good and thus must be fought for without actually knowing why. This is dangerous, as this is how cults start. The Cult of Decentralization.
Decentralizing for the sake of Decentralization
Ask any ‘decentralist’ what it means to be ‘decentralized’ or how they measure it, and they will fumble. They will offer excuses as “it is hard to measure clearly” or “you can’t measure it directly, but you can measure certain aspects of it”. Which, to any trained skeptic is just poppycock self-rationalizations of an assumed belief system which they do not actually understand, but believe via faith alone.
They will say that more nodes makes a network more decentralized (not really, if all the nodes are in one area, or controlled by one person). Or they will say then that more independent nodes controlled by different people are better than a few controlled by one, for that is obvious. Well, if you think it is obvious, then you have fallen into the mind trap. The mistake is to stop here because this rationale is based purely on the technical definition of the word “decentralized”. Indeed topographically, in graph theory and mathematics, the notion of more connections between nodes makes a system less centralized and more decentralized, ending in the maximally decentralized topology “distributed”. But if you end the discussion here, then you have fallen prey to the magicians trick. The classic distraction of the audience to mask the true illusion, which is of course, that the REAL thing we care about is censorship resistance, and network security. Somehow, after all the brouhaha and cries for decentralization, most people will have forgotten to ask the most important question of all… is decentralization the same as Security? If not, and it is only a proxy for security, then it stands to reason that we must try to show by what measure or relation does increasing decentralization actually increase security. Is it a linear relationship? Is it monotonically increasing? Is it just positively correlated? In which domains? Should we be also evaluating other solutions which may not increase decentralization, but actually increase security? In fact, the whole “Maximum Decentralization is the goal of Bitcoin” narrative of the Cult of Decentralization breaks down, along with the core foundation of the ‘small blocker’ camp’s arguments.
So how DO you measure decentralization? Well, that is a foolish question, if you are measuring the wrong thing. The correct answer is how do you measure security, and how much does decentralization affect it. I have mentioned this many times on other occasions but network security cannot be measured by theoretical technical analysis alone. In adversarial game theory, you must always consider the socio-economic costs of an attack. Thus, the only way to measure security, (and by extension how much decentralization matters or affects it) is in terms of dollars, or what I call the “Cost of Corruption”. Think of it as the cheapest possible method by which an attacker could compromise the network to do something which is against the rules (or social contract) of the network. You can see this as an event where a government wishes to block a users payments, confiscate property unlawfully, or even if a bunch of golden-hearted bandits decide to steal from the rich and give to the poor. In Bitcoin, one way you can empirically discover one cost of corruption is by how much it would take to bribe 51% of the hash power of miners to block payments from an individual or party.
Now the reason why it is very important to think of Bitcoin network security in terms of Cost of Corruption is because then you start to appreciate where the TRUE dangers to network security may be, and spoiler alert, the number of individual nodes in the network isn’t the biggest worry. For instance in order to actually get 51% of the hash power to collude to defraud or act against the network, you have to work out what they are losing in return. They risk losing all future revenue streams from their business, as well as socio-economic losses in personal reputation. This means that a small mining operation is cheaper to bribe than a larger one, because they have less expected future revenue. Furthermore, they have less of a potential loss due to less capital invested into their business. This means that the constant point repeated by decentralists that we need more diversity in mining nodes is a red herring, unless you further qualify that those node collectively must be more costly to bribe than 1 large mining node. (a hundred 16-year olds running mining nodes on their laptops can be pretty cheap to buy off via anonymous bounty than 1 large business with future revenue streams to protect). This is also pertinent, with respect to the much debated safe activation threshold for a hard fork (with 60%, 75%, 90%, or 99% of the majority). Traditionally, a supermajority in politics involves a 67% majority. This is no accident. There are good reasons for this, but unfortunately many of the computer scientists who have little background in socio-economic systems discount it, calling it a relic of governments and Bitcoin is a new system that need not play by old rules. This is just ignorance.
The reason why 67% is statistically significant is because it is the ratio which makes the Cost of Corruption the most expensive, and thus, the least likely to be corrupted. Why? Consider if the majority rule were to be triggered at 90%. This means that a group controlling 11% of the vote would be able to block any measure or proposal from passing. This may be seen as a good thing from the perspective that the ‘minority should not suffer by the majority’, but this situation, called Minoritarianism, is actually quite susceptible to corruption due to the fact that an attacker need only to bribe, threaten the business, or kidnap the loved ones of 11% of the vote in order to control the system. Having compromised 11%, the attacker can then demand and influence which proposals the majority proposes (and vow to veto all others). Considering the bottom 11% of hash power in the Bitcoin network are likely small mining operations with equivalently small capital investments, they make easy blackmail and extortion targets, and are more likely to give in to attacker demands. The level of 67% is significant because, it sets the Cost of Corruption at 33%, which is the highest it can be while still having the majority support be greater than 2x that of the minority; any more and the majority would no longer be a supermajority.
So knowing this, you can start thinking critically about what is the lowest cost to corrupt the network today? Where do the biggest risks lie? I leave you with this challenge: Which do you think is cheaper, to bribe one or more developers to mislead 51% of miners into installing malicious code, or to have 51% of the miners collude amongst themselves to create and install malicious code*?
Next time anyone says that this or that is good for decentralization, you know what to ask them, to see if they really know what they are trying to sell you, or if they are just repeating the common in-vogue party line.
*Hint: one solution involves considering the mining businesses future expected revenue streams, the other does not.