Foreword: Is only 1% of honest nodes feasible in reality?
"If you add more assumptions, specifically, this requires observers to actively monitor consensus, not just download the output afterwards, then you can increase the fault tolerance rate to 99%."
In a highly technical article commented by Emin Gun Sirer, V God suggested that this means that more than 50% of miners or rights pledges are not required to be honest, and only 1% are honest.
- How does Vitalik evaluate Ethereum Eco-Privacy, DeFi and Ethereum 2.0?
- Vitalik Buterin: Not very optimistic about the alliance chain, 5G technology can improve blockchain scalability
- Sudden! V God turns on the AMA answer mode to see what kind of "difficulty" he has encountered.
- Chinese New Year Reading Notes | Is Vitalik's Focus on Cryptoeconomics Solved Now?
- When the funds are exhausted, the Ethereum Foundation will become history
- Exclusive dialogue V God! Where is the Ethereum Road under question?
Therefore, the frightening 51% attack is 99% attack in the case of this consensus algorithm, which means that the attack actually becomes impossible.
From a high-level perspective, how it works, as we understand it, works by requiring a listening node or an independent observer.
You may be familiar with the rules first seen in Bitcoin. If you are not familiar with this, this is the explanation of Nakamoto Satoshi on July 17, 2010. We fully quoted it:
“I believe that for a payment processing company, it is possible to provide a fast distribution transaction service and be able to complete sufficient inspections in 10 seconds or less.”
The network node only receives the first transaction version they received in order to merge into the block they are trying to generate. When you broadcast a deal, if others broadcast a double flower transaction at the same time, this is about the contest of who first broadcasts to most nodes. If a person starts a little ahead of time, it will spread faster through the network and will reach the majority of the nodes first.
A rough example:
This means that if a double flower transaction has to wait even if it is only a second, then it has a huge disadvantage.
The payment processor has connections to many nodes. When it receives the transaction, it will spread it vigorously while monitoring the double transaction of the network. If it receives a double flower transaction on any of the listening nodes, it will warn that the transaction is bad. If no listener hears it, the double flower transaction will not go too far.
The double flower transaction will have to wait until the end of the listening phase, but by that time, the payment processor's broadcast has arrived at most of the nodes, or it is far ahead in terms of broadcasting, and there is no hope that the double flower transaction can be surpassed.
Buterin proposed something similar, but for the block. Buterin later clarified that, as he clearly stated in the original proposal, he did not invent the 99% fault-tolerant consensus algorithm, which was invented by Leslie Lamport. Lamport is a computer scientist who won the Turing Award for his contributions to the field of distributed systems.
“I just wrote an interpreter and adapted it into a blockchain context,” says Buterin. In the high-level interpretation of Ethereum developer Conrad Barski, he pointed out:
“Vitalik suggests that if an independent observer of network traffic (that is, a blockchain client running by the user, not a miner/verifier) observes what is happening in real time and pays attention to when the information appears, they can monitor the miners’ initiation of 51. The "fouling behavior" of % attacks, which provides additional security and can prevent such attacks."
Buterin later added: "Another use case is that it can be used as a tool to detect 51% of attacks, and coordinate the use of a few soft forks to get rid of them without having to rely on too many human-driven social coordination tools to Select a behavior or something else."
This sounds like fraud proof, some bitcoin core developers say such mysterious creatures do not exist. However, Nakamoto's white paper mentions such fraudulent proofs to explain why there is no need to retain specific data, or why light nodes can be very secure.
Simply put, as the name suggests, fraud is evidence of fraud, deception or misconduct, or evidence that there is no consensus.
If you run the full node, you are verifying everything, so the node will reject the misbehavior. However, if you are not verifying everything, and you are using a smartphone, then in rare cases you may receive incorrect data and you will not know that they are incorrect.
Fraud proofs provide this approach. The advice here is that if we understand very well, it seems that the independent observer is monitoring, so you can know that the data is incorrect.
Of course, all of this is done by the software or the code itself. If you "know" some data in the code is incorrect, it will automatically reject it. Buterin said:
"If 5% of the verifiers are honest, then none of the 512 randomly selected nodes are honest, and the probability is only about one in a trillion. Therefore, as long as the network delay plus the clock skew is less than D/ 2, the above algorithm will be effective, it can correctly coordinate the nodes on some single final value, even if there are multiple conflicts in the final value (because the fault tolerance of the threshold correlation algorithm is destroyed) can be solved correctly."
The focus here is of course Casper, and this proposal seems to be a recommendation to include it in PoS. If it works in this context, then the Ethereum blockchain may become more secure.
The author of this article, Trustnodes, is translated by the Blue Fox Note Community "LS".
Appendix: Vitalik's 99% Fault Tolerance Consensus Guide