Are Miners Centralized? A Look into Mining Pools

Read about Alethio’s analysis of the (de)centralization of the Ethereum network. A week’s worth of data from the Ethereum blockchain provides transparency and insight to mining pools and miners.

by Danning Sui, Saulo Ricci and Johannes Pfeffer

Ethereum is backed by p2p network architecture, in which each participant is a node. Each node in this distributed network is connected with other nodes [1,2], and all nodes must reach a consensus to maintain the continuity of the blockchain. Some nodes, called miners, invest their computing power in order to confirm the next block to be appended to the blockchain.

Oftentimes, miners will combine their hashpower to increase the likelihood of successfully mining a block together and having more frequent payouts. These conglomerates are known as mining pools, and can wield significant influence over a blockchain. Current academic sources judge the (de)centralization of the Ethereum network by the hash power of mining pools [4,5,7]. We believe this view of oversimplified and instead provide nuance to the discussion. This article first takes a look at the current state of mining pools and summarizes their arithmetical power. Additionally, by analyzing patterns of how mining pools pay their miners, we gather deeper insight into the distribution of mining power on the Ethereum network among actual miners.

Different Shades of Mining

There are three primary methods of mining for a regular individual [3,6]:

1. By oneself: A miner can mine by oneself. Though she would receive all the rewards without having to pay anyone else, the approach is usually impractical in reality for casual miners with a mining rig or two. The difficulty in popular blockchains like Ethereum has become so high that finding a block would take years. She would be competing with big mining pools and corporations that own and deploy massive amounts of energy and computing power. The upside is that she would have full control over all aspects of the mining (such as what transactions to include).
2. Joining a mining pool: A miner could join a mining pool. By adding her hash power to the power of the group, she increases the pool’s chances to verify a block. When anyone in the pool finds a valid block, the reward is shared among the miners proportional to the work they have submitted.
3. Joining a cloud mining: Instead of owning the hardware required to mine herself (either individually or in a pool), the miner could rent time from existing miners who operate the hardware. Though she pays a fee to the third-party miners, she avoids the financial overhead of purchasing and maintaining mining hardware.

Our Dataset

The dataset used in our article encompasses a week of blocks and contains transactions of the Ethereum blockchain from 04/21/2018 to 04/28/2018*.

Table 1: Dataset Summary

Are Mining Pools Centralized? Yes.

To give grounds for our analysis, we begin by demonstrating how mining power is distributed among all mining pools. As illustrated in Figure 1, the top five pools mined 84% of all the newly-found blocks within the week analyzed. At a high level, mining pools are, in fact, quite centralized. A mining pool, however, is quite granular in reality. A pool includes the interests of all individuals who invested their hardware with the aim of increasing their hashing power. Our analysis dove beyond the mining pool level to investigate just how decentralized Ethereum truly is.

Figure 1: Pie Chart of number of mined blocks in the time frame. The chart illustrates how each mining pool contributed with its hashpower to mine a week of blocks.

Are Miners Centralized? Not Really.

At first glance, the graph above suggests that mining power is quite centralized. However, as mentioned before, mining pools don’t directly own huge mining machines, but rather apply lots of individual computational power to mine blocks. When we investigate the computational power of individual miners, therefore, we see a different picture.

The most transparent way mining pools reward their participants is through direct payouts. In this case, there are direct transactions from the address of the mining pool to the actual miners. These payments are usually recurring and their amount can be set by the miner in the pool’s configuration interface. Sometimes the payouts are not made directly from the pool’s address but from separate address (called proxies). There are some pools that use a much less transparent payout scheme: they directly pay the miners in fiat currency. One example for this can be cloud mining, where the contributor pays the cloud rent via credit card. Our analysis found different patterns of payouts.

Table 2 shows the top miners over the seven days analyzed. The payouts should be roughly proportional to the number of blocks mined. There can be some distortion as payout heights are custom and time delayed. For the last pool in the list, bitclubpool, the payout is clearly not proportional to the number of blocks. Bitclubpool is one of the pools that don’t transfer the block rewards directly to the miners (see the number of unique payout accounts). This makes their payout scheme much more opaque and reduces trust, in our opinion.

Table 2: Summary of the mining pools who contributed the most with their hashing power during the week examined**

Figure 2 depicts the Cumulative Distribution Functions (CDF) for the percentage of total rewards of a pool paid out to each miner. A CDF graph depicts the distribution of the sample among values. Steep slope means sample are centered around this value marked in x-axis, with y-axis showing the percentage of samples that are equal to or under this value. The CDF charts show that small payouts make up the vast majority of payouts from pools to actual miners. We found that the four strongest pools on average pay less than 0.04% to each miner. The maximum value found was for a participant from nanopool, who was able to receive 3.7% of all rewards from the network.

Figure 2: Cumulative Distribution Functions of the payout percentage destined for each miner for the top 4 mining pools — ethermine, f2pool, ethfans, and nanopool

How are Mining Pools Paying Miners? We Found Various Patterns.

We have observed three different ways a mining pool relays rewards to its miners:

Paying directly on chain with cryptocurrencies (e.g. ethermine, ethfans, f2pool, nanopool, etc.)
1. Small value of payout to lots of payout accounts
2. Mostly paying less than 1 Ether in each transaction to various accounts
3. Payout accounts are actual miners

Paying via proxy on chain (e.g. Address 0xb75d1e62b10e4ba91315c4aa3facc536f8a922f5)
1. Large payouts to a few accounts
2. Paying around 100 Ether in each transaction, to a specific account
3. Payout account is a proxy, and the proxy account pays to actual miners

Paying off chain, e.g. via credit card (e.g. bitclubpool)
1. Large Ether transfers to a few accounts
2. Roughly paying over 1000 Ether in each transaction, to some specific accounts
3. Payout accounts are likely to be the pool owners’ wallets.

Figure 3 shows a scatter plot for each mining pool demonstrating how many beneficiaries (miners) got paid out against the average of their payments. It points out the different payout schemes and helps in clustering mining pools accordingly.

Figure 3: Scatter plot for each mining pool of the number of beneficiaries against the average paid out*** (both axes are in logarithmic scale)

We can see that the top five mining pools, miningpoolhub, f2pool, ethfans, nanopool and ethermine, share a similar pattern: large number of payout accounts and small average payout value. This is consistent with the decentralized structure of miners in the pool, and confirms they are paying their members directly on chain.

On the other hand, bitclubpool stands out as an outlier on the top left, showing an extremely high average payout going to only a few accounts. These accounts are not receiving payouts from block mining work, as 35,173 ETH is way beyond the rewards gained for the 1124 blocks mined by the pool during the week. The bitclubpool website**** claims they pay out miners with fiat money via checking accounts. We believe the large on-chain payouts from bitclubpool are most likely a transfer of value to other pools’ owners. Beside, there are indicatives of scamming activities that should be obfuscated by the mining activity from bitclubpool and bw [8,9].

Furthermore, we checked the other two data points near bitclubpool and bw. For address 0x92e3f585ab699445fd7a809b5edf0c3dc88fb549, the payout pattern is similar to bitclubpool and bw.

For address 0xb75d1e62b10e4ba91315c4aa3facc536f8a922f5 on Figure 3, we observed that the payouts were around 100 ETH each, and all went to only one account. This account then paid small amounts (<1 ETH) to various other accounts, which is a pattern similar to that of top mining pools who have a payout proxy.

How Many Miners Would be Needed for a 51% Attack? Several Thousand.

The question of a 51% attack frequently circulates in the blockchain ecosystem. When an attacker controls more than half of the networks hashrate she always creates the longest chain over time, even if she never includes blocks from other miners [10,11]. Since the longest chain is always accepted as the authoritative one, the attacker now controls:

What transactions are included in the blocks (censorship).
Reversal of shorter competing chains (double spend and other attacks).
Many more negative effects.

We gathered all the actual miners across the different mining pools and ranked them by the rewards they gained throughout the week, which we assume to be proportional to the mining power they provide.

In order to achieve 50+ percent of the mining power, an attacker has to convince at least 4484 miners to cooperate — in other words: at least 4484 miners would need to act maliciously to overthrow the Ethereum network, not just 3 pools. However, 4484 top miners make up only ~1.5% of all actual miners (pool payout addresses) in the network. Figure 4 illustrates that 297,443 actual miners control the other 50% of the mining power on the network. An attacker, therefore, would only need to compromise 1.5% of the network — the most powerful nodes — to control the network.

Figure 4: Pie chart of the percentage paid out for each actual miner during the week. ( The brown half is actually made up by 4484 colorful pieces, but due to the resolution limit, it looks like a solid brown piece )

A successful 51% attack is likely to have a very negative effect on the market value of a cryptocurrency. A long-lasting 51% attack could even stop them from mining valid blocks at all if all competing chains would be completely reversed by the attacker. Assuming rational behaviour of the miners, they want to protect the value of their investment and their earnings. If a single pool would reach a hashrate close to 50% or visibly colludes with other pools to mount a 51% attack, the miners would abandon these pools to protect their income. Additionally, for the above reasons, a 51% attack by a legitimate pool (a pool that exists to make money by mining) contradicts its raison d’être.

Looking Down the Road

The state of the Ethereum network is ever-changing. Previously, we have not had the tools to effectively and dynamically monitor the health of the network and the miner ecosystem to inform the different user groups, e.g. the miners, about patterns of activity that affect them. The team at Alethio has built a product that allows us to do just that. With Alethio, the Ethereum network is available to anyone to analyze, study, and draw conclusions. Learn more about us at https://aleth.io/.

Notes

* All the values related with payouts and rewards are addressed in ETH.

** As a simplification, we are considering the miners those ones who got paid out directly from the pool and that number is stated on the last column of Table 2. Note, this approximation is a lower bound and the number of actually miners might be even bigger.

*** Beneficiaries is considered as all miners who got paid out at some point during the week considered.

**** https://bitclubpool.com/

******An actual miner is defined as an account who submits work to a mining pool and has received direct payouts from the pool. The set of actual miners is derived as follows:

The miner has received in average less than 500 ETH per transaction from the mining pool. The threshold of 500 ETH was empirically suggested.
This heuristic is of course flawed (not all payout addresses are individual entities, not all pools payout to the actual miners directly).

***** The following steps were followed in order to get the set of actual miners who dominated 50% of the mining power:

1. Sorted the list of all actual miners in descending order by the percentage of the payout received from the overall payout paid by all miners during the week. Lets call this feature as avg_percentage_payout;
2. Returned the sliced list of actual miners up to 50th percentile of the cumulative value of avg_percentage_payout.

References

1. The Meaning of Decentralization, URL: https://medium.com/@VitalikButerin/the-meaning-of-decentralization-a0c92b76a274, accessed on 05/02/18
2. Questions on terms ‘distributed’ and ‘decentralised’, URL: https://ethereum.stackexchange.com/questions/7812/question-on-the-terms-distributed-and-decentralised, accessed on 05/02/18
3. How to mine Ethereum, URL: https://www.coindesk.com/information/how-to-mine-ethereum/, accessed on 05/02/18
4. Decentralization on Bitcoin and Ethereum, URL: http://hackingdistributed.com/2018/01/15/decentralization-bitcoin-ethereum/, accessed on 05/02/18
5. Are we decentralized yet?, URL: https://arewedecentralizedyet.com, accessed on 05/02/18
6. Ethereum Mining: the ultimate guide on how to mine Ethereum, URL: https://www.bitdegree.org/tutorials/ethereum-mining/, accessed on 05/02/18
7. Ethereum mining faces “a centralization problem”, URL: http://bitcoinist.com/ethereum-mining-faces-centralization-problem/, accessed on 05/02/18
8. [Investment Warning] Bitclub Network Review and Analysis, URL: https://99bitcoins.com/anatomy-bitcoin-scam-bitclub-network-analyzed/, accessed on 05/02/2018
9. Why is BW mining pool no longer listed as a top miner?, URL: https://www.reddit.com/r/ethereum/comments/55zjae/why_is_bw_mining_pool_no_longer_listed_as_a_top/, accessed on 05/02/2018
10. Weaknesses: Attacker has a lot of computing power, URL: https://en.bitcoin.it/wiki/Weaknesses#Attacker_has_a_lot_of_computing_power, accessed on 05/02/2018
11. 51% Attack — Investopedia, URL: https://www.investopedia.com/terms/1/51-attack.asp, accessed on 05/04/2018

Disclaimer: The views expressed by the author above do not necessarily represent the views of Consensys AG. ConsenSys is a decentralized community with ConsenSys Media being a platform for members to freely express their diverse ideas and perspectives. To learn more about ConsenSys and Ethereum, please visit our website.