As academic paper titles go this is pretty normal.
> Can someone summarizes to me the genius behind the idea?
I'm reading quickly, but I think the idea is that the reason we need proof-of-work is to prevent Sybil attacks, that is, a single entity pretending to be 50% of the network via spoofed identities. So you have a concept of persistent identity for each (apparent) member of the network, and you require nodes to solve a computational problem when they join the network and also periodically while they're in the network. This puts relatively little computational load on each participant, but puts a lot of load on a long-term attacker, and even more work on a short-term attacker who's trying to claim a bunch of identities in a hurry.
I'm not sure how much this actually helps Bitcoin, since my impression is that the computational load is what's needed to match the abilities of the legitimate members of the network. I guess the trick is that maybe you can make the block-mining difficulty scale up less aggressively over time, but I'm not following that logic yet.
Thinking harder I guess the only reason that mining difficulty is so competitive is that there's a reward for mining, given to just the miner, so the arms race is profitable (much as, say, the arms race for microsecond-level tech improvements is profitable to HFT because even if two algorithms make the same decision only one gets the purchase in first).
So if you get rid of that, and maybe add a nominal "reward" for solving the puzzles to remain in the network and give it to all participants, the incentive to build giant mining farms goes away: the only reason to have additional computational power is either to keep up with the network as a whole (not the fastest person in the network) or to actually gain 50% of computational power.
As academic paper titles go this is pretty normal.
> Can someone summarizes to me the genius behind the idea?
I'm reading quickly, but I think the idea is that the reason we need proof-of-work is to prevent Sybil attacks, that is, a single entity pretending to be 50% of the network via spoofed identities. So you have a concept of persistent identity for each (apparent) member of the network, and you require nodes to solve a computational problem when they join the network and also periodically while they're in the network. This puts relatively little computational load on each participant, but puts a lot of load on a long-term attacker, and even more work on a short-term attacker who's trying to claim a bunch of identities in a hurry.
I'm not sure how much this actually helps Bitcoin, since my impression is that the computational load is what's needed to match the abilities of the legitimate members of the network. I guess the trick is that maybe you can make the block-mining difficulty scale up less aggressively over time, but I'm not following that logic yet.