To avoid the centralization of network power, how does "reverse incentive" work?
The concentration of power faced by various blockchain networks has been a topic of extensive discussion among researchers and token holders. In fact, this is also an important issue in the eyes of blockchain designers and developers. They hope to decentralize power as much as possible in the design and fit the original intention of a decentralized network.
secondary title
A question about "redistribution"
The issue of centralization is mainly about two aspects: how tokens are distributed and redistributed. The former is "token economics" - what kind of initial network power can be obtained for what kind of contribution; the latter is "staking economics" - how the obtained power generates new benefits, and what rules are used to reward and punish.
secondary title
"Reverse incentives" that cannot be ignored
image description
Source: Cosmos Hub official website
In the Cosmos Hub mainnet genesis file, we can see three aspects of current slashing in this parameter:
Temporary bans due to poor availability (long offline time);
The proportion of penalties for poor usability, 0.01%;
The proportion of double signature penalty, 5%;
articlearticleThere have been relevant introductions.
In June of this year, the first double-signature confiscation occurred on Cosmos, resulting in a total loss of $60,000 in equivalent tokens for nodes and clients. Although on the Tezos network, slashing is only for verification nodes, and a long list of records about network penalties also makes people aware of the necessity of setting up and selecting a reliable node.
secondary title
"Proportionate slashing" broadens the reverse incentive effect
Cosmos core developer Sunny Aggarwal recently released a draft of ADR (Architecture Decision Records) that improves slashing, proposing proportional slashing, discussing how to make slashing not only affect network security, but also try to influence the game of participants strategy, which in turn makes nodes tend to be smaller and more dispersed. It is pointed out in the plan that the current trend of centralization on the Cosmos network is not good for the network: it will increase the risk of censorship, and may cause a decline in activity and fork attacks, which is a negative externality.
And due to the existence of "sybil attack (referring to a node in the network actively operating multiple identities at the same time)", there will always be some obstacles in the design attempt to avoid centralization. In other words, the same node operator can use the rules of the protocol to create different identities to offset the actual effect of decentralized participation brought about by redundant storage in the blockchain. For example, split a large node into many small nodes.
On the draft, Sunny Aggarwal said:
We propose to implement a "proportional slashing" proposal to counter Sybil attacks. That is to say, in this network system, instead of imposing the same degree of penalty on each node according to a single percentage, it implements different penalty ratios according to the network power owned by a verification node. Sunny uses progressive formulas to show the step-by-step thinking process:
（1）slash_amount = power
// power is the faulting validator's voting power.
（2）slash_amount = (power_1 + power_2 + ... + power_n)
// where power_i is the voting power of the ith validator faulting in the period.
（3）slash_amount = (sqrt(power_1) + sqrt(power_2) + ... + sqrt(power_n))^2
（4）slash_amount = k * (sqrt(power_1) + sqrt(power_2) + ... + sqrt(power_n))^2
// where k is an on-chain parameter for this specific slash type
To sum it up, it’s actually more intuitive to say——
Assuming that a node has 10% of network power, then we gradually add conditions, and finally form the slashing formula determined by (4):
(1) The amount of confiscation is determined according to the voting power of the network, and the confiscation is set to 10% (temporarily assumed to be the same as the network power);
(2) What if it is split into small nodes? According to the correlation, the voting rights of nodes that make mistakes at the same time within a period of time are added together to determine the network penalty ratio. Taking two splits into two as an example, the penalty ratio is 5%+5%=10%, and there will be no less penalty.
(3) In order to further suppress splitting nodes and make the degree of penalty for splitting be higher than the actual proportion of network power, then on the basis of (2), calculate the square of the sum of the square roots of the network power of each small node. In this way, the slashing ratio of two nodes with 5% network power will rise sharply from 10% to 20%.
(4) Determine different K values for different types of mistakes, and then multiply K by the result of (3) to distinguish different degrees of punishment for different mistakes.
secondary title
Promoting Diverse Staking
If this new penalty mechanism is followed, operators with higher network power may be punished in greater proportion. Just as Sunny Aggarwal said, “Two factors, the proportion of voting power of nodes that make mistakes within a certain period of time and the number of nodes that make mistakes at the same time for a period of time, can jointly affect the slashing standard.”
secondary title
Plans to be refined
It is undeniable that the design of the confiscation system does help to improve the "Gini coefficient", make the "decentralization" of technology a step further, and promote a more secure and stable blockchain network.
Some community members agreed and raised questions and suggestions, such as: Are the criteria for determining relevance too broad? Preventing "sybil attacks" is a good idea, but is it too harsh and should another parameter be set? Some people reiterated the importance of setting the upper and lower boundaries of the penalty ratio, and advocated setting a reasonable upper limit to avoid excessive punishment. Since this system may make the cost of doing evil for small nodes lower, it is necessary to seriously consider the minimum penalty value.
The voice of concern is that although it is a good idea, it is believed that token trustees do not realize that this is a risk at all, and only focus on profits. There is also a view that the expected effect is based on the assumption that the verification nodes are very rational, but in fact it is not always the case. It is hoped that the scheme can first change the strategy of some rational verification nodes. At the same time, various tools This difference in forfeiture ratio should also be reminded.
There is an argument that this is just a "placebo" for centralization. The reason is that the zero transaction fee of some large nodes is so attractive that people only care about choosing a free and convenient solution regardless of the future.
secondary title
More Thoughts on Centralization
People's views on network centralization come from many aspects, and they hope to truly decentralize network power through system design. In May of this year, the verification node Chainflow announced an organization called Decentralized Staking Defenders, which aims to change the trend of oligarchs that may appear on the network and jointly maintain a "more balanced, fair, and joinable" ecology. And it seems that the discussion about distribution and redistribution is not a technical issue, but a social topic.
The topics of society and governance are more difficult issues than technology, as Vitalik once stated. For example, we can achieve "equality" through various means, but how to achieve "equity" may depend on the individual.
References:
References:
