Tobo
|
|
November 16, 2015, 12:40:31 PM |
|
If quantum computers become popular, what impact will it bring to Bitcoin and other existing cryptocoins?
|
|
|
|
patmast3r
|
|
November 16, 2015, 12:48:08 PM |
|
If quantum computers become popular, what impact will it bring to Bitcoin and other existing cryptocoins? Pretty much all current public-key cryptography is done for.
|
|
|
|
|
Come-from-Beyond (OP)
Legendary
Offline
Activity: 2142
Merit: 1010
Newbie
|
|
November 16, 2015, 02:06:06 PM |
|
Pretty much all current public-key cryptography is done for.
PoW blockchain mining too. PoS/PoI/DPoS blockchains unaffected.
|
|
|
|
Tobo
|
|
November 16, 2015, 02:25:14 PM |
|
Pretty much all current public-key cryptography is done for.
PoW blockchain mining too. PoS/PoI/DPoS blockchains unaffected. How difficult is it for Bitcoin and Ethereum to switch their current algorithms to quantum resistant cryptographic algorithms?
|
|
|
|
Come-from-Beyond (OP)
Legendary
Offline
Activity: 2142
Merit: 1010
Newbie
|
|
November 16, 2015, 02:26:22 PM |
|
How difficult is it for Bitcoin and Ethereum to switch their current algorithms to quantum resistant cryptographic algorithms?
Ethereum is migrating to PoS. Bitcoin will be dead.
|
|
|
|
patmast3r
|
|
November 16, 2015, 02:30:19 PM |
|
Pretty much all current public-key cryptography is done for.
PoW blockchain mining too. PoS/PoI/DPoS blockchains unaffected. I guess those would be unaffected but those projects would still need to swtich from ecdsa to something else that is "quantum secure" for all their signing right ?
|
|
|
|
Tobo
|
|
November 16, 2015, 02:30:39 PM |
|
How difficult is it for Bitcoin and Ethereum to switch their current algorithms to quantum resistant cryptographic algorithms?
Ethereum is migrating to PoS. Bitcoin will be dead. But Ethereum still needs to take care their public key attack issues. Is it easy for Etheruem or POS to switch to a new algorithms to protect the public keys?
|
|
|
|
Come-from-Beyond (OP)
Legendary
Offline
Activity: 2142
Merit: 1010
Newbie
|
|
November 16, 2015, 02:33:23 PM |
|
I guess those would be unaffected but those projects would still need to swtich from ecdsa to something else that is "quantum secure" for all their signing right ?
Don't know about the others, but Nxt Account Control feature includes QC-resistance.
|
|
|
|
Come-from-Beyond (OP)
Legendary
Offline
Activity: 2142
Merit: 1010
Newbie
|
|
November 16, 2015, 02:36:20 PM |
|
Is it easy for Etheruem or POS to switch to a new algorithms to protect the public keys?
Existing schemes of QC-proof signing require pretty big signatures. In Iota, for example, ~433 bytes are occupied by the signature, it's exactly 50% of the transaction size.
|
|
|
|
tromp
Legendary
Offline
Activity: 990
Merit: 1108
|
|
November 16, 2015, 02:39:29 PM |
|
Pretty much all current public-key cryptography is done for.
PoW blockchain mining too. Hashcash PoW (like Bitcoin's and most altcoin's) is amenable to Grover search which can search a space of n nonces in time O(sqrt(n)). But Hashcash with large memory requirements will likely not be affected as long as scaling quantum computers up to millions of bits remains elusive. Non-hashcash PoWs like Cuckoo Cycle are even less affected, as they are immune to Grover search.
|
|
|
|
Come-from-Beyond (OP)
Legendary
Offline
Activity: 2142
Merit: 1010
Newbie
|
|
November 16, 2015, 02:49:54 PM |
|
Hashcash PoW (like Bitcoin's and most altcoin's) is amenable to Grover search which can search a space of n nonces in time O(sqrt(n)).
But Hashcash with large memory requirements will likely not be affected as long as scaling quantum computers up to millions of bits remains elusive.
Non-hashcash PoWs like Cuckoo Cycle are even less affected, as they are immune to Grover search.
"Improvements" like Scrypt make nodes more vulnerable to spam. Also, a quantum computer doesn't need to evaluate a whole hash value, it can verify first bits and throw away nonces that don't suit with high probability. BTW, why is Cuckoo Cycle less affected, birthday paradox problems are solved with N^(1/3) effort VS N^(1/2).
|
|
|
|
tromp
Legendary
Offline
Activity: 990
Merit: 1108
|
|
November 16, 2015, 03:39:36 PM Last edit: November 16, 2015, 03:50:56 PM by tromp |
|
Hashcash PoW (like Bitcoin's and most altcoin's) is amenable to Grover search which can search a space of n nonces in time O(sqrt(n)).
But Hashcash with large memory requirements will likely not be affected as long as scaling quantum computers up to millions of bits remains elusive.
Non-hashcash PoWs like Cuckoo Cycle are even less affected, as they are immune to Grover search.
Also, a quantum computer doesn't need to evaluate a whole hash value, it can verify first bits and throw away nonces that don't suit with high probability. Computing a single bit of a hash is almost as much effort as computing the whole hash; you might be saving a percent or two at most. BTW, why is Cuckoo Cycle less affected, birthday paradox problems are solved with N^(1/3) effort VS N^(1/2).
Because, unlike Birthday collision problems, Cuckoo Cycle is a more structured search problem; you must find a 42 cycle in an arbitrary graph. There are no known quantum speedups for such graph problems.
|
|
|
|
Come-from-Beyond (OP)
Legendary
Offline
Activity: 2142
Merit: 1010
Newbie
|
|
November 16, 2015, 03:51:03 PM |
|
Computing a single bit of a hash is almost as much effort as computing the whole hash; you might be saving a percent or two at most.
Could you provide a proof of this statement? http://jheusser.github.io/2013/02/03/satcoin.html claims the opposite.
|
|
|
|
Come-from-Beyond (OP)
Legendary
Offline
Activity: 2142
Merit: 1010
Newbie
|
|
November 16, 2015, 04:28:53 PM |
|
But Hashcash with large memory requirements will likely not be affected as long as scaling quantum computers up to millions of bits remains elusive.
I didn't find information on time-memory trade-off of quantum computers, but if we assume that the trade-off is not worse than the trade-off of classical computers then we get that memory increase of the hashing function can be counteracted by increasing time we run the computations. So Hashcash with large memory won't save us.
|
|
|
|
tromp
Legendary
Offline
Activity: 990
Merit: 1108
|
|
November 16, 2015, 04:33:59 PM |
|
Computing a single bit of a hash is almost as much effort as computing the whole hash; you might be saving a percent or two at most.
Could you provide a proof of this statement? Thus follows directly from how SHA256 is defined. It is many rounds of confusion and dispersion; so that each single bits in one round depends on pretty much all bits of previous rounds. That is a long document to read. Where exactly does it claim that?
|
|
|
|
tromp
Legendary
Offline
Activity: 990
Merit: 1108
|
|
November 16, 2015, 04:37:39 PM |
|
But Hashcash with large memory requirements will likely not be affected as long as scaling quantum computers up to millions of bits remains elusive.
I didn't find information on time-memory trade-off of quantum computers, but if we assume that the trade-off is not worse than the trade-off of classical computers then we get that memory increase of the hashing function can be counteracted by increasing time we run the computations. So Hashcash with large memory won't save us. Of course I was talking about hash-functions that don't allow for time-memory trade-offs.
|
|
|
|
Come-from-Beyond (OP)
Legendary
Offline
Activity: 2142
Merit: 1010
Newbie
|
|
November 16, 2015, 05:21:04 PM |
|
Of course I was talking about hash-functions that don't allow for time-memory trade-offs.
Give me the name of one of such functions, please. The trade-off is a pretty universal thing, the best a function can do is to keep time*memory*advice constant, if I'm not mistaken.
|
|
|
|
Come-from-Beyond (OP)
Legendary
Offline
Activity: 2142
Merit: 1010
Newbie
|
|
November 16, 2015, 05:27:47 PM |
|
so that each single bits in one round depends on pretty much all bits of previous rounds.
This means that after some number of rounds SHA256 doesn't give a better mixing, hence it's possible to do a shortcut by finding a polynomial with fewer number of operators. That is a long document to read. Where exactly does it claim that?
I introduced a novel algorithm to solve the bitcoin mining problem without using (explicit) brute force. Instead, the nonce search is encoded as a decision problem and solved by a SAT solver in such a way that a satisfiable instance contains a valid nonce. The key ingredients in the algorithm are a non-deterministic nonce and the ability to take advantage of the known structure of a valid hash using assume statements.
A couple of benchmarks demonstrated that already with simple parameter tuning dramatic speed ups can be achieved. Additionally, I explored the contentious claim that the algorithm might get more efficient with increasing bitcoin difficulty. Initial tests showed that block 218430 with considerably higher difficulty is solved more efficiently than the genesis block 0 for a given nonce range. This means that in average computation of a single bit takes less time than computation of the whole hash.
|
|
|
|
tromp
Legendary
Offline
Activity: 990
Merit: 1108
|
|
November 16, 2015, 06:25:11 PM |
|
Of course I was talking about hash-functions that don't allow for time-memory trade-offs.
Give me the name of one of such functions, please. The trade-off is a pretty universal thing, the best a function can do is to keep time*memory*advice constant, if I'm not mistaken. You are quite mistaken. This is a recognized weakness in scypt's design. Here's one: Argon2, winner of the Password Hashing Competition. Most of the PHC candidates qualify, since time-memory-trade-off resistance was one of the design goals.
|
|
|
|
|