/tech/ - Tech


[New Post]

(93.2KB, 1600x718)
Because SHA3 is complete and utter garbage.
If I have to recommend a hash algorithm, I would choose BLAKE2, or if people want to have something fresh, BMW, SHABAL and skein (in decreasing order of speed).
Even though ECHO and Shavite3 is fast (faster than BLAKE2, as fast as BMW) they rely on AES instruction sets which makes them useless in ceratin cases.
I would never recommend CubeHash (not even CubeHash832) as the person making it was a fucking troll.
Replies: >>30 >>41 >>58
(1.4MB, 1646x2142)
Take a look at this
Replies: >>28 >>42
Pretty cool. I guess that is some scanned paper from a competition of all the hashes? I know there is a well known one, just forgot the name.
I wonder how long until blake2 starts getting better adoption over sha2. Imageboards for example could use it for file hashes where speed is pretty important.
Replies: >>31
>>26 (OP) 
Why do you think so about SHA-3? Its main concept (sponge) is really cool.
Replies: >>31 >>32
The numbers are "cycles per byte", the lower the number the faster the speed.

Looks good but slightly hard to implement, the original is slow but SHAKE, KangarooTwelve and MarsupilamiFourteen is fast, BLAKE2 is way more standardized.
>>30 I'm more interested in tree hashes. I see one marked on the page. Not sure when they first became a thing (not a crypto guy) but the idea seems good for torrenting.
Replies: >>34
Tree Hash as a speed-up makes no sense, but MerkleDAG from IPFS is promising as FUCK
Replies: >>34
(144.5KB, 1437x885)
>>33 Tree hash isnt meant for a speedup. Its partly meant for verifying parts before the whole, which is why >>32 mentioned torrenting and partly to help torrent file size.
You get the root node hash and then peers send data from branches. The hashes of concatenated hash of the branches should hash into a parent node. So the torrent file would only need to include the root hash. The rest you get from peers.
As torrents get bigger, there would need to be either:
1. more hashes of blocks in torrent file s to prevent fake blocks (bigger torrent file)
2. larger blocks (more possibility of bad data before hash checked)
You can read more here: http://www.bittorrent.org/beps/bep_0030.html
Wiki page also describes what I was saying in pic related
No people do use tree hashes for "speed-ups" in practice (see Keccak's tree hashing)
BUT the Merkle Tree idea for Torrent is very similar to the idea of MerkleDAG in IPFS, which is pretty good
>>26 (OP) 
IIRC Sha-2 is fastest because of hardware implementations.
Replies: >>42
(22KB, 528x352)
AES has hardware implementations and judging by the table in >>27 some hashing algorithms are using it. This is above my head as far as /tech/ goes but isn't AES for encryption and hash algorithms are supposed to be one-way only?
Replies: >>43
AES as scrambling function my boi, scrambling and compression functions have different uses within the MD construction.
>>26 (OP) 
>Because SHA3 is complete and utter garbage.
why though
Replies: >>198
speed, speed, speed!
Replies: >>199
I mean if its faster and still has all the good properties against collisions and stuff, then it becomes more useful as a checksum rather than password hashing. Since there is already bcrypt and pbkdf2, SHA is becoming less used for passwords.
Not bad overall IMO.
Replies: >>238
KangarooTwelve > SHA3

15 replies | 4 files
Show Post Actions


Staff Actions:


- rules - faq - source code -