What are the different types of 51% attacks?

There are different types of 51% attacks, and although they can achieve the same attack effect, they are fundamentally different for network participants. From the perspective of network participants, there is a 51% attack (internal computing power attack) that has nothing to do with the overall computing power of the network, and the network can only reduce the risk of being attacked by shortening the block time.

Proof of Work

Proof of Work is the cornerstone of the network's security, and its main principle is that network participants independently verify transactions by consuming a large amount of energy. Based on the average energy consumption of the network (hash rate), network participants decide how many blocks they will wait to be sure that the transaction is protected - for example, if a participant decides to wait for K blocks and the average computing power consumed by these K blocks is X, then he believes that the transaction will remain on the chain after spending K * X computing power. After K transaction confirmations, network participants are confident that the probability of malicious actors generating K * X computing power to deceive network participants is very low and conclude that the transaction is immutable.

We can imagine a hashing ball that ensures that we don't have to trust each other. The hashing balls can have different colors (the network uses different hash functions), and the hashing ball can only compete with another hashing ball if it has the same color. The larger the hashing ball, the more secure the network. This idea is great and works very well.

External computing power attack

A classic example of a 51% attack is when a malicious actor buys a large amount of hardware and starts mining in secret. A simpler approach is to rent hashrate through a service like NiceHash. In both cases, we have a hashrate sphere of size E that protects the network, and an external attacker needs to generate a new hashrate sphere larger than E for K blocks (the number of blocks required to reverse a transaction).

As a participant in the network, it is relatively easy to protect yourself from these attacks. Just wait for enough transaction confirmations. It is worth noting that networks that are more vulnerable to such attacks are those with a smaller overall hashrate, where an attacker can obtain more than 51% of the total network hashrate through hashrate rental services. Another case is if another network hashrate ball has the same color (i.e. the same hash function) and the network hashrate ball is larger (such as Bitcoin and Bitcoin Cash). To prevent attacks from another known hashrate ball, an alarm system can be set up to check if these hashrate balls are getting smaller - the disappearing hashrate can be used to attack other networks.

Internal computing power attack

The hashrate that protects our network can be divided into multiple smaller hashrates that compete with each other. For example, if we have a network where 65% of the hashrate comes from only 4 different sources (such as mining pools). If the internal hashrate of the network is used to attack it, then we have a completely different situation from the perspective of the participants.

51% attack using internal hashrate

The hashrate attack in the above figure looks the same as an external hashrate attack, but the difference is that the probability of transaction confirmation is no longer a function of the security provided by the hashrate, but only a function of time. To make this clearer, let's imagine a network protected by a large hashrate (it accounts for 90% of the total hashrate on the planet). An external attack in this case is impossible. However, for an internal hashrate attack, this situation is no different from the case where the hashrate accounts for 0.0001% of the total hashrate, because this attack uses a certain percentage of the existing hashrate. In this case, network participants can only rely on the time that has passed since the transaction was included in the block (block time), and the size of the hashrate cannot protect against this type of attack. This is why these two cases are different types of 51% attacks. Internal hashrate attacks may be easier to detect on networks with shorter block times, because we can see the latest hashrate distribution in a shorter period of time to respond accordingly to various situations (such as some hashrate pools suddenly stopped mining).

Internal + external computing power attack

A hybrid of the two attacks above is also possible. The attacker uses less than 51% of the internal hashrate (e.g. 40%), which is combined with a new hashrate that did not exist before (20%). In this case, the size of the network's own hashrate is sufficient to protect against external hashrate, so this attack scenario is not as bad as a pure internal hashrate attack.

Bitcoin hashrate distribution

We don’t know if these attacks are possible on the Bitcoin network. There may be some limitations on the Bitcoin network that make internal attacks almost impossible. Since mining pools need to pay miners very frequently, they may not have enough funds to support them, so the implementation of a 51% attack may not ultimately be profitable for them.

The current Bitcoin computing power distribution is as follows:

The fact that the majority of the Bitcoin network’s hashrate is in the hands of more than 10 mining pools does not seem optimal, increasing the risk of an insider attack. Even if all 10 hashrate sources are controlled by honest network participants, their systems are at risk of being hacked, and if they use the same client, there is also a risk of vulnerabilities being exploited. In the end, this article still believes that Bitcoin is the most secure network for transferring value with the least amount of trust, but it does seem like there is room for improvement.

Summarize

From the perspective of a network participant, there is a type of 51% attack (an insider attack) where participants cannot use the size of the hashing ball to protect themselves. This is because the size of the hashing ball has nothing to do with an insider attack. For this attack, participants can only rely on the time that has passed since the transaction was included in a block (block time). It is worth noting that networks that are able to produce blocks quickly have an advantage in this situation because they can get more information about the losing pool in a shorter time. Insider attacks deserve some attention, and research needs to be conducted to determine if it is possible to distribute hashing power in a better way.