The malicious actor could then go back in the blockchain to the block before the BTC transfer was confirmed and mine an alternate chain, in which the BTC transfer is not included. The majority share of the networking power would ensure that alternate is forced on the rest of the network.
While an attacker can cause a great deal of damage when conducting a 51% attack, a majority attack does not allow the malicious actor to prevent transactions from being broadcast, nor does it enable them to reverse transactions from other users. Changing the block’s reward, creating coins out of thin air, or stealing coins that never belonged to the attacker are also very improbable scenarios.
The further back a transaction is, the harder it would be to subvert it, as the number of new blocks to be mined to bring the network up to the current block height becomes further and further away. This is the reason why Bitcoin transactions usually require a threshold of 6 confirmations before clearing.
A 51% attack on the Bitcoin blockchain is very unlikely because of the magnitude of the network. As the network grows, the possibility of a single person or entity obtaining enough computing power to overwhelm all the other participants gets increasingly unrealistic.
Blockchains prevent a 51% attack using a combination of techniques. One way is by encouraging more participants to join the network and run their own nodes to support the network. The larger the number of participants contributing their resources, the more difficult it becomes for a single entity to dominate the network.
Consensus mechanisms, such as Proof-of-Work (PoW) and Proof-of-Stake (PoS), also play a crucial role in preventing 51% attacks. These mechanisms require nodes to agree on the validity of transactions before adding them to the blockchain. They make it economically and computationally expensive for an attacker to manipulate the network, as they would need to control a majority of the network's resources.
Promoting decentralization is another preventive measure. Distributing nodes across different locations and involving various participants means no single entity gains excessive control.