All serious research into POS has shown that "Nothing at Stake" is the bogey man that POW/Bitcoin Maximalists use in in fear tactics.
This may have been possible for the earliest POS coins, but newer, solid implementations have nothing to fear. That is what the research says (5-6 full research papers with statistical models are available for those who want to go neck deep into the specifics):
To summarize the discussion, known claimed attacks on proof-of-stake distributed consensus algorithm(and concrete implementations) at the moment:
1.
Short-range attack - attacker can offer better chain started few blocks behind current canonical chain. The attack is possible at the moment, the only likely outcome though is just gathered fees increase for an attacker. In our simulations this kind of attack is possible mostly when a long delay occurs due to low target. By the way, the attack has positive aspect for network, as it shorten delays average between blocks. So attacker gets extra fees for a good job done
2.
Long-range attack - attacker can start fork hundreds or thousands blocks behind current chain. From our investigations the attack isn't possible.
3.
Nothing-at-stake attack - not possible at the moment! Will be possible when a lot of forgers will use multiple-branch forging to increase profits. Then attacker can contribute to all the chains(some of them e.g. containing a transaction) then start to contribute to one chain only behind the best(containing no transaction) making it winner. Previous statements on N@S attack made with assumption it costs nothing to contribute to an each fork possible and that makes N@S attack a disaster. In fact, it's not possible at all to contribute to each fork possible, as number of forks growing exponentially with time. So the only strategy for a multibranch forger is to contribute to N best forks. In such scenario attack is possible only within short-range e.g. with 25 confirmations needed 10% attacker can't make an attack. And attack is pretty random in nature, it's impossible to predict whether 2 forks will be within N best forks(from exponentially growing set) for k confirmations. So from our point of view the importance of the attack is pretty overblown.
4.
History attack - attacker can buy whale's private key for $5 and build alternative story. Solved with some checkpoints now, located behind max rollback possible, so the solution is not so scary in terms of centralization etc.
If you know any other kind of attack, please add. Please note IPO properties of a concrete coins etc isn't related to proof-of-stake distributed consensus problems.
And Consensus Research is going to work on better proof-of-stake prototyping & implementation !
This research is specifically into Nxt's POS mechanism. If unsure about your POS coin, ask your dev what POS implementation it is based on.
