In centralized computer environment, errors are tightly monitored and swiftly corrected, so reaching consensus is fast and reliable. A decentralized computer environment, however, is much more vulnerable to failures and attacks, thus requiring much stronger reliability and security protocols. Ethan Buchman, co-founder and CTO of Cosmos Tendermint, notes that a distributed consensus mechanism such as Casper must “make a reliable system from unreliable parts.”
Dynamic validator sets as security
Ethereum PoS must solve the difficult problem of ensuring ‘consensus’ among participants so that all remote processes arrive at the same conclusion. In solving this problem, Casper makes Ethereum PoS security stronger than Ethereum PoW security by algorithmically leveraging ‘stake’ in the form of bonded ETH. Stake that is bonded by a node which exhibits faulty behavior will be partially or completely forfeited or ‘slashed’ by Casper, with the penalty for incorrect behavior far outweighing any transaction fees or mining rewards defined for correct behavior.
Validators have less incentive to maliciously attack something in which they have a personal stake. Casper is programmed to systematically destroy an attacker’s Ether as soon as malicious activity is recognized. All validators have known identities established by stable Ethereum addresses and the network keeps track of the size and makeup of each validator set.
The are two ways of automating social authentication to reduce the load on users:
1. Take BIP 70-style payment request in recent block hash, validator makes sure that they are on the same chain as the vendor before approving a payment or any on-chain interaction.
2. Use universal hash time (UHT) so successful attack chain must be generated secretly at same time as legitimate chain was being built, requiring majority of validators to secretly collude for relatively long time.
Security vs safety (robustness vs performance)
‘Security’ addresses intelligent threats, i.e., a node is “right” or “wrong”. ‘Safety’ addresses the inherent problems of a decentralized computer network like power outages or messaging errors; the node is “up” or “down”. Sometimes security measures weaken safety and vice versa.
Blockchain Safety vs Liveness
PoW and PoS blockchain “safety” is a consensus protocol’s ability to ensures that every transaction is replicated and recorded in all clients in the network in the same order, where there are not two or more competing chains with valid transactions. The blockchain is ‘safe’ when all blocks have either “settlement finality” or “probabilistic finality”.
PoW and PoS blockchain ‘liveness’ is availability for, and responsive to, new transaction propagation, where valid transactions eventually make it onto the blockchain. A liveness fault occurs a transaction is omitted, withheld, or reordered. Liveness faults are indistinguishable from simple network latency and, therefore, difficult to detect.
Denial-of-service attack security
“Spawn camping” is a denial-of-service cyberattack where a malicious POW miner with more than 50% of network hash power repeatedly send bogus transactions to network, prevent processing legitimate transactions, thus rendering network useless. A malicious Casper PoS attacker cannot spawn camp without destroying their own Ether every time. Beyond that disincentive is the fact that repeated large-scale attacks would reduce Ether supply, increasing the cost of buying ETH, and making an attack more expensive each time its executed.