Classical Proof of Work systems are becoming more and more threatened by 51% attacks. This is caused by the combination of at least two factors. One is that cryptocurrencies are now expensive enough to be attacked. The other factor is that PoW mining is becoming increasingly centralized, which is partly because of specialized hardware, like ASICs. In this centralized landscape, one could even rent some hashing power to attack a system. Here is a very informative site demonstrating this. It is likely that those systems that do not come up with an idea for protection against 51% attacks will be wiped out soon. Here we list some alternatives for the future, the information was gathered from the forums of Dash, Decred and ZenCash. Thanks for the community of these projects for their help!
Dash has a two-tier system, which consists of miners, as in PoW systems, and MasterNodes, who are special nodes staking 1000 Dash. MasterNodes are requared for InstantSend and PrivateSend. However, a normal transaction is not affected by MasterNodes, a normal transaction is handled by the PoW system. In the future, Dash will probably switch to so called collateralized mining.
What is collateralized mining?
A collateralized miner is a combined MasterNode and miner. In the future, Dash MasterNodes will require a complex set of specific hardware, including ASICs.
Step by step mining without proportionate ownership of MasterNodes will be prohibited, after that, mining will require so called "mining keys". This will lower energy consumption, as there is an upper limit how much one can mine. The collateralized mining will greatly reduce the chance of a 51% attack and contribute to the scalability of the network. It is supposed that this way the network could handle 50 MB blocks easily.
Another thing that might reduce the chance of a 51% attack is the BetterHashMiningProtocol update, where instead of pools, miners decide what will go into the next block, making the system more decentralized. This is not a planned update in Dash, but the idea was presented on the forum.
The system of Decred is a hybrid PoW/PoS. A 51% attack against Decred would be ~20 times more costly that of a classic PoW system, let's say Bitcoin. Here is an article describing this. Decred was based on memcoin2, but it also has elements from Proof of Activity.
Proof of Activity
Proof of Activity is a Hybrid PoW/PoS consensus algorithm, where miners only create empty block headers, which is then passed to 3 randomly selected stakers. The stakers need to be online, which is enhancing decentralization, because these will be full nodes. If some of the randomly selected PoS node is not online, the process needs to start over, a new block header needs to be mined. If they are online, they verify the block header. The last "lucky stakeholder" (3rd) adds the transactions to the block, signes it and broadcasts it. The transaction fees are shared among the miner who found the block header and the N-lucky-stakeholders (N is advised to be 3)
memcoin2
In mc2 the randomness to select stakeholders who will sign the next block is achieved with a ticket system. After the stakholder submitted some coins to the system to purchase a ticket, ~23 days need to pass to obtain a voting ticket. Randomness is based on hashes from the block headers, as in PoA. This way, one ticket holder will be selected (she wins the "lottery"). That ticket holder will be allowed to sign the next so called stake sub-block. Stake sub-blocks extend PoW blocks. The purpose of stake sub-blocks is to validate the block that was created by miners.
Decred
So, Decred is mainly based on memcoin2, but it implemented some elements from PoA as well. Indeed, it created it's own consensus mechanism, namely Decred, from the starting point of memcoin2 and PoA. These other projects were published back in 2013 and 2014, so we can see that the idea of alternative consensus mechanisms to combat 51% attacks are not that much new.
Decred is supposedly well protected against a 51% attack.