Mutable and immutable blockchains

Now we all know what the bitcoin blockchain is, don’t we? It’s just one particular version of the general class of blockchains, which share the characteristics that data is stored in blocks and because of some cryptographic jiggery-pokery the blocks are chained together, so that you can’t go back and change the contents of a block without having to then change the contents of every subsequent block. And depending on the consensus protocol that is used, you can’t change the blocks without everyone else agreeing to let you do it. Thus it is, as my colleague Salome Parulava describes it, “mutable by consensus”.

Whereas auditing at present entails the confirmation of transactions and balances on a company’s accounting ledger at the end of the period, a transaction on the blockchain would provide a permanent and immutable record of the transaction almost immediately.

From Blockchain and the Auditing Revolution – Real Time Audit within the Capabilities of Blockchain | Fintech Schweiz Digital Finance News – FintechNewsCH

The reason that this kind of structure is called immutable, even though it is mutable by consensus, is that it is computationally infeasible to go back post-consensus and make a change. Even if you obtain consensus and co-ordinate more than half of the “hashing power” in the case of bitcoin, and could in theory go back to the very first block, change it to send the bitcoins in it to yourself, and then go forward rewriting all of the subsequent blocks, it would take years and years of massive computing power. Someone could, in theory, treat all of the bitcoin transactions from the last checkpoint up until now as the wrong side of a fork. (For all we know, secret mining pools are As my good friend Gideon Greenspan pointed out to me, just because you could see that corrupt agents were rewriting history in this way it doesn’t mean that you could stop them. But it’s not a realistic attack. We can live with the description “immutable” to mean “theoretically mutable but not mutable under any practical circumstances that we can envisage”.

If you had a different kind of blockchain, however, you could design it work in a different way. It could be mutable by consensus, or mutable by a dictator, and it could be mutable in a computationally feasible way. This is what some researchers in the US and Italy have put forward in a paper called “Redactable Blockchain, or Rewriting History in Bitcoin and Friends” (5th August 2016). Giuseppe Ateniese, Bernado Magri, Daniele Venturi and Ewerton Andrade say: 

We put forward a new framework that makes it possible to re-write and/or compress the content of any number of blocks in decentralized services exploiting the blockchain technology. As we argue, there are several reasons to prefer an editable blockchain, spanning from the necessity to remove improper content and the possibility to support applications requiring re-writable storage, to “the right to be forgotten”.

We don’t need to go into the clever mathematics behind this. Just take forward the idea that you can use that clever mathematics to substitute for massive amounts of computing power that I mentioned above and can rewrite a block without having to go forward and rewrite all subsequent blocks. The well-known and well-respected outsourcing company Accenture has filed a patent on this idea with Professor Ateniese.

By allowing a central administrator to amend or delete information stored on a blockchain, the [outsorucing company, Accenture] says that its prototype will make the technology more attractive to the financial services industry.

From Accenture to unveil blockchain editing technique – FT.com

This announcement was met with widespread derision on social media, and I can understand why. One of the key reasons for considering a blockchain to implement certain kinds of financial services is that the state of the blockchain, the shared world view, is locked down and the end of each block. If the shared world view can be changed, it wouldn’t be useful for these services any more. Now, I can see why some people might want an accounting system that works this way (see, for example, the case of Kingfisher Airlines in India) but I wouldn’t have thought that society wants accounting systems that work this way at all. Why would you want a ledger that can be edited either by some group or subgroup of the consensus forming stakeholders or by some central authority? I can think of a few reasons, but none of them make any sense.

The financial services industry needs to face the question of how to balance the appeal of pristine accounting with the demands of the real world, where some things simply need to be struck from the records.

From Downside of Bitcoin: A Ledger That Can’t Be Corrected – The New York Times

Nothing ever needs to be “struck from the records”. If a bank makes a mistake — let’s say it accidentally opens a couple of million bogus accounts — then it can’t just go back and scrub the backup tapes and pretend it never happened. The way to correct a wrong debit is with a correct credit. The Financial Times quotes blockchain entrepreneur and serious player Blythe Masters, the former JPMorgan banker running Digital Asset Holdings, as saying of Accenture’s approach that “we think it is innovative and can strike the right balance between preserving blockchain’s key features and adapting it for real-world requirements within some permissioned systems.” But what are these real-world requirements within some permissioned systems that Ms. Masters is referring to?

I don’t think anyone would use the bitcoin blockchain consensus protocol that was designed for an open, permissionless  blockchain (i.e., proof of work) for a closed, permissioned blockchain so you would never need to edit it this way. My reading of the paper, from a not-a-cryptographer perspective, is that it does not deliver against the real-world requirements for permissioned systems in financial markets. The use cases that are set out in the paper are the need to remove child pornography from a public blockchain, the “right to be forgotten” and the need to consolidate records financial transactions. My feeling is that none of these are real-world requirements.

As for the first use case, this is not something that our clients need consider since none of them are proposing to implement critical national financial infrastructure on a public blockchain with arbitrary content controlled by unaccountable consensus groups. If, for example, a stock exchange were to implement a blockchain settlement system, it would not be of such a type as to allow members of the general public to store child pornography on it (or at least it wouldn’t be if it had people such as Consult Hyperion designing it).

What’s more, if a stock exchange were implemented on an editable blockchain, it would be utterly chaotic since at the execution of any transaction, no-one could be certain about the state of the ledger (since it would be possible for some future intervention to change it). My granny dies and leaves me IBM shares. I sell you my IBM shares. I use the money to buy a car. Then a decade later a court order overturns my granny’s will as it turns out she had a son that we’d never heard of. So we go back and change the blockchain so that the IBM shares belong to him instead of me. So now I didn’t have the money to buy the car. So I have to give the car back. But the car was scrapped… and so on. Interstellar overdrive… then I go back five years later because it turns out he wasn’t her son at all and now I want the blockchain changed to give me my IBM shares…

Richard Lumb, global head of financial services at Accenture, told the Financial Times that financial institutions and regulators would need a means to quickly correct errors on the blockchain before using it in securities markets. He gave the example of a “fat finger” trading error, or a trade assigned to the wrong counterparty.

From Accenture to unveil blockchain editing technique – FT.com

That’s not how you correct errors, by just rubbing out mistakes. These are regulated financial institutions, not the mafia. No-one is going to build a financial services market on top of a mutable blockchain. In one of the comments I saw about this proposal, someone said that it would be OK because the market participants would keep an audit log of the changes and who agreed them. I thought that perhaps such an important log might need to be stored on an immutable ledger. Uh oh, blockchain Inception. 

As for the next use case, I am not a lawyer, but I think that the paper misinterprets the so-called “right to be forgotten”. However misguided the European Court’s decision on this might be, it does not demand the rewriting of history. If you publish an article about me that I think contains “old, inaccurate or even just irrelevant data“, and I manage to persuade Google that it should be harder to find, then the article is not deleted. The link to the article is removed from Google search results but the article is still there. Here, for example, is the Daily Telegraph’s full list of stories that have been removed from search results.

Newspapers are not required to go back and tear out articles from their archives, they are exempt (but in Europe, Google opted not to be regulated as media company so is not exempt). And I’m sure none of us what would to live in a world where politicians could obtain court orders to go back a change the historical record! When it comes to the serious use cases (e.g., revenge porn) it is already impossible to purge the matrix and it won’t make any difference whether they are stored on a blockchain or not (although with a permissioned blockchain you would at least know who had put them there and therefore who to arrest).

The third use case, the consolidation of financial records is not clear to me at all. Since the invention of double-entry bookkeeping, the whole point of keeping a ledger has been that you have a record of all of the credits and debits that contribute to the current world view. Companies do not delete old transactions every few months to save space. In fact the law requires them to maintain the transaction records for years. Here’s one example: in the UK, the “direct debit guarantee” has no time limit at all, so all records relating to direct debits need to be kept forever. If there is something about this use case that I haven’t understood, I would be genuinely interested to be corrected.

In summary, then. We all appreciate the clever mathematical tricks behind the mutable blockchain, but when it comes to the serious world of banking and financial services, it seems like (in the casual demotic of our unlearned age) a bit of a chocolate teapot.