Let’s talk about digital currency mining again: Are Ethereum miners centralized?

Yesterday, ChainNews published an article titled "Understanding the Current Situation and Secrets of Cryptocurrency Mining Industry in One Article", which comprehensively analyzed the nature of mining and mining machine business. Today, we focus our attention on the most popular Ethereum mining field recently.

As ChainNews reported before, Bitmain's Ethereum mining machine has been released, and several other large mining machine manufacturers have also disclosed plans to manufacture Ethereum mining machines, which makes some miners worry that this will easily lead to the centralization of Ethereum mining and the risk of 51% attacks by manufacturers with a large market share. Alethio team members used data on Ethereum to deeply analyze the current status of Ethereum mining. However, although the team concluded that "it is difficult to achieve the attacks that some miners are worried about in the Ethereum network anyway", one thing that cannot be ignored is that an attacker only needs to unify 1.5% of the most powerful nodes in the entire network to control the entire Ethereum network.

Author: Danning Sui, Saulo Ricci and Johannes Pfeffer

Compiled by: Zhou Hengyu

The essence behind Ethereum is a p2p network structure, and each Ethereum participant is a node in this huge structure. In this distributed network, each node is connected to other nodes, and in order to maintain the sustainable development of the entire block network, all nodes need to reach a consensus. Some of the nodes actually belong to miners, whose job is to continuously contribute computing power to confirm the next block to be put on the chain.

In order to increase the success rate of mining and obtain more frequent returns, miners often cooperate with each other to pool their hash power for mining activities. This combination of hash power is what we know as a mining pool. The huge computing power of the mining pool will have an impact on the entire blockchain network.

Recent academic research has used the hash power of each mining pool as the criterion for judging the centralization of Ethereum, but we believe that this method of judgment not only oversimplifies the problem, but also brings some differences to the discussion of the centralization of computing power.

We hope to provide readers with a deeper understanding of the current operating status of various mining pools and their computing power capabilities. In addition, we will also analyze the way mining pools pay miners to gain deeper understanding of the distribution of computing power on Ethereum.

Let me first state our conclusion. In the Ethereum ecosystem:

• Are mining pools centralized? The answer to this question is: “Yes”, mining pools are highly centralized organizations.

• Are miners centralized? Actually not;

• Furthermore, in the Ethereum network, the attacks that some miners are worried about are difficult to implement anyway;

• However, an attacker only needs to unify 1.5% of the most powerful nodes in the entire network to control the entire network.

Let’s take a look at our detailed analysis.

What are the methods of mining?

Let's start with the most basic knowledge. For regular individual miners, there are generally three ways to mine:

1. Mining by yourself

Miners can organize their own computing power to mine. Although miners can enjoy all the benefits in this way, for amateur miners with only one or two mining machines, this method is not feasible in reality, because in these popular blockchain networks like Ethereum, it may take several years to find a block using this weak computing power.

These miners have to face the competition of computing power from large mining pools and cooperative groups with a lot of electricity resources and computing power resources. This approach has only one advantage, that is, miners can master all the processes in mining.

2. Join a mining pool

Miners can also join a mining pool to mine. By adding their own hash power to the mining pool's hash power, miners increase the possibility of verifying a block. Once a miner in the mining pool discovers and verifies a block, miners in the entire mining pool share the mining revenue in proportion to their hash power contribution.

3. Cloud Mining

In addition to owning their own hardware to mine, including self-mining and mining pool mining, miners can also rent idle machines from miners who own their own mining machines. Although they need to pay fees to third parties, in this process, miners who use cloud mining avoid investing too much in the initial cost of purchasing and maintaining mining machines.

The data we used in our research

The database used by our institute contains Ethereum block data for the entire week from April 21 to 28, 2018, including Ethereum transfer data, etc.

Overall status of the dataset

Conclusion 1:

Are Ethereum mining pools centralized? Yes

To support our analysis, let’s first talk about how computing power is distributed among all mining pools.

As shown in the figure below, the top five mining pools mined 84% of the newly confirmed blocks this week. So in a sense, mining pools are highly centralized organizations.

But in the real world, each mining pool is a relatively independent entity. A mining pool will include the interests of all individual miners who invest their own mining equipment in order to improve their mining capabilities. Therefore, in our analysis, we will go deep into the mining pool to prove how Ethereum achieves the goal of decentralization.

The proportion of blocks mined by Ethereum mining pools

Conclusion 2:

Are Ethereum miners centralized? Actually not

The first impression given by the above graph is that the computing power of Ethereum mining is very concentrated. However, as mentioned above, mining pools do not essentially own a large number of mining machines. They actually mine new blocks by combining the computing power of many individuals.

When we look at this matter from the perspective of individual miners, we will find that the whole thing is not what we imagined.

The most direct way for a mining pool to pay miners is to return the proceeds directly. In this case, the proceeds are directly transferred from the mining pool account to the miner's account address. This distribution is a commonly used distribution method, and the distribution amount can be set by the miner who controls the mining pool distribution account.

In some cases, the payment of such rewards will not be sent directly from the mining pool account, but will be returned to the individual miners who participated in the mining through a separate account called a proxy account. However, some mining pools use less direct payment methods: they directly use legal currency to pay the rewards to individual miners. For example, the cloud mining we mentioned earlier pays the rewards in this way, and they will deposit the rewards into the credit cards of the participants in the entire mining process.

The table below lists the top mining pools during the period we analyzed. Their mining revenue is roughly proportional to the number of blocks they mined. The discrepancies in the data mainly come from the personalization of the payment process and time delays.

We can see that the last mining pool in the table, bitclubpool, pays a disproportionate amount of rewards to the mined blocks. Bitclubpool is what we call a mining pool that does not directly return rewards to miners from the mining pool account. We can infer this from the number of independent payment accounts. We believe that this way of operating rewards is very opaque and will reduce miners' trust in it.

The image below depicts the cumulative distribution function (CDF) of the percentage of rewards paid by the mining pool to each miner. The image of the cumulative distribution function mainly illustrates the distribution of samples. The steep curve indicates that the area near this value on the x-axis is a relatively concentrated area of ​​samples, and the y-axis shows the proportion of samples below this value in the total sample.

The cumulative distribution graph shows that small payments to miners account for the vast majority of all payments. Our analysis found that the top four mining pools only pay an average of 0.04% of their total rewards to each miner. The miner who received the largest proportion of the total reward was a participant in "nanopool", who received 3.7% of the overall mining pool revenue.

How do mining pools pay miners?

During our analysis, we observed three different ways of paying rewards among mining pools:

1. Use cryptocurrency to pay directly on the chain. Mining pools that do this include Ethermine, Ethfan, F2pool, nanopool, etc.

Features:

• Each payment amount is small, and there are many payment accounts

• Most payments are less than 1 ETH

• The account owners who make payments are the miners who actually participate in mining.

2. Pay the remuneration through the proxy account on the chain. For example, the following account is a proxy account used for payment on the chain 0xb75d1e62b10e4ba91315c4aa3facc536f8a922f5

Features:

• For large payments to a small number of accounts

• On average, each transfer pays about 100 ETH to a specific account.

• The account is a proxy account, and the payment is made directly to the miners who actually participate

• Through non-chain accounts, such as credit card payments, such as Bitclubpool

• Large amount transfer payment

• The amount is roughly 1000 ETH per transaction

• The payment account is usually the personal account of the mining pool owner

The following figure is a scatter plot of the number of miners who received payments and the average payment amount. It includes different payment methods and makes corresponding clusters.

We can find that the top five mining pools, miningpoolhub, f2pool, ethfans, nanopool and ethermine, all have the same pattern: a large number of miners and small payments. This means that they pay miners directly, which is consistent with the decentralized structure of miners in the mining pool.

But from another perspective, "bitclubpool" is an outlier at the top left of the graph, which means that "bitclubpool" only pays a large amount of rewards to a few accounts. These accounts do not receive their rewards from the block mining network, because we can obviously see that the 35,173 ETH they received far exceeds the income of the 1124 blocks they mined that week. Bitclubpool's website says that they use checking accounts to pay rewards through fiat currency. We believe that a large number of on-chain payments on Bitclubpool are actually transfers to other mining pool owners. At the same time, there are signs that Bitclubpool and bw are using mining to cover up certain fraudulent activities.

In addition, we carefully explored the points near Bitclubpool and bw in the above figure. The payment pattern under the address 0x92e3f585ab699445fd7a809b5edf0c3dc88fb549 is similar to Bitclubpool and bw.

The mining pool payment at the address 0xb75d1e62b10e4ba91315c4aa3facc536f8a922f5 in the above picture also has certain characteristics. We found that each payment is worth about 100 ETH, and all are transferred to the same account. Then this account will transfer the income to other accounts with an amount less than 1 ETH, which is somewhat similar to the practice of the top mining pools using proxy accounts.

How many miners are needed to launch a 51% attack? Thousands

The 51% hashrate attack is a common problem in the blockchain ecosystem. When an attacker controls more than half of the hashrate of the entire network, he can eventually create the longest chain in the entire network. This chain of recorded information belongs entirely to him and may not even contain any blocks previously mined by other miners. Because the longest chain is always considered to be the one with the best credibility, the attacker has the following privileges:

• Control which transfers are authenticated by the blockchain Censorship

• Cancelling a slightly shorter competing chain can cause double spending and other attacks

• Causes many negative effects

We collected information about miners from different mining pools and sorted them according to the rewards they received during the week of the survey. Based on the ratio of computing power to revenue, we can determine the amount of computing power they control.

In order to achieve more than half of the computing power, an attacker must try to convince more than 4484 miners to cooperate before launching an attack; that is, at least 4484 miners must unify malicious attacks to subvert the entire Ethereum network, rather than the three mining pools that can subvert the network. However, the top 4484 miners only account for 1.5% of the total miners in the network calculated by payment addresses. The figure below shows that 297,443 miners control the remaining 50% of the computing power. Therefore, from another perspective, an attacker only needs to unify 1.5% of the most powerful nodes in the entire network to control the entire network.

Once a 51% attack is achieved, it will have a huge impact on the entire cryptocurrency market. If the competing chain is not as long as the attacking chain, a sustained 51% attack will even prevent other miners from mining new blocks from the network. Assuming that miners are rational enough, they will try their best to protect the value of their investment and returns. If a mining pool reaches nearly 50% of the hash power of the entire network, or cooperates with other mining pools to reach more than 51% of the hash power, miners will abandon these mining pools in order to protect their own profits.

The author of this article is a member of the Alethio team. Alethio is a tool that monitors the health of the Ethereum network and miner ecosystem, providing different users with information that affects their network activity patterns. Through Alethio, everyone can analyze some of the conditions of the Ethereum network and draw some conclusions through research.