:celeste_hearts_lesbian: :celeste_hearts_trans: :celeste_hearts_demisexual: flori_ava_star@[matrix]conf:~:cursor_blinking: (@star)

flori_ava_star@[matrix]conf:~

@star@amazonawaws.com

if you use MD5 or SHA1 for literally ANYTHING in 2026, STOP. GET SOME HELP. DO NOT

1❤️4

Yureka ✨ @yuka@fedi.yuka.dev

@star I use sha1 all the time, I’m sure you do too looks at the git repos 🙃

flori_ava_star@[matrix]conf:~

@star@amazonawaws.com

@yuka ACTUALLY git 3.0 will make sha256 the default iirc :3

ciara @doti@pl.catboyindustries.co

@star @yuka perhaps this is a stupid question but isn’t blake way more efficient is there a reason people aren’t using it(?)

flori_ava_star@[matrix]conf:~

@star@amazonawaws.com

@doti @yuka blake is a much faster hashing algorithm, yes. But I guess sha is much more widely known, adopted and its security profile has thus also been evaluated much more thoroughly than blakes, and for something like git where this choice you make now has to still be respected 10, 15, 20 years from now, that might be more important

ciara @doti@pl.catboyindustries.co

@star @yuka yea that’s kinda what i was thinking. maybe in 2048

flori_ava_star@[matrix]conf:~

@star@amazonawaws.com

@doti @yuka i use blake3 in projects where i want a fast hashing algorithm! ^-^

flori_ava_star@[matrix]conf:~

@star@amazonawaws.com

even for seemingly inane tasks, simply DON'T!!!!

❤️2

Xerz 💗 @xerz@masfloss.net

@star sometimes, the only available hash for a file is MD5 and then I

❤️1

[4fbb09] @5225225@furry.engineer

@xerz @star to be fair, a lot of the cases where a file hash is provided (foo.net listing the hash of a file you download... from foo.net) is simply not security relevant.

unless the file is actually being hosted on a mirror and the file hash isn't.

in that case, yeah, it makes sense (but also you still kinda need to trust the mirror because in 99% of cases people won't check the hash unless you do it automatically)

❤️1

flori_ava_star@[matrix]conf:~

@star@amazonawaws.com

@5225225 @xerz noooooo, "not security relevant" is not a good excuseeee!!!! this is how usage of md5 proliferates/continues lkjhslkjhdfglkjhsdflkjhsdfglkjhsdgflkjhsdgflkjh

there are simply much better, much safer, much faster hashing algorithms out there! like sha512 or blake3 my beloved

❤️1

Juan D. Jara :yeah_waffy: @javascript@app.wafrn.net

@5225225 @star @xerz

when someone sends you a md5 hash in the year 2026

[4fbb09] @5225225@furry.engineer

@star @xerz i mean for the case of file checksums my argument is to not provide a md5 it's to not provide a hash at all. no one checks it, and checking it only protects against disk corruption on your end (which, sure. whatever.), or the file getting corrupted on the server end.

Xerz 💗 @xerz@masfloss.net

@5225225 @star ...huh? I check hashes constantly tho, in particular for Windows software and retro stuff, plus my Nix checks hashes constantly in an automated fashion, just as e.g. torrents do

this seems like a really easy way to lose track of history and force everyone to be vulnerable just because the majority doesn't understand hashes

[4fbb09] @5225225@furry.engineer

@xerz @star the specific case i'm talking about is page on https://foo.net/ with a download link to https://foo.net/file.bin, and a hash (secure or otherwise) on that page

the hash there is pure security theatre. anyone who can tamper with the file can tamper with the hash to match their tampering. okay, there's more vague benefits of "it's easier to check that the hash hasn't been tampered with than the file" (if anyone is actually monitoring that...), but for the individual, it's not as meaningful

for the case of nix, yes, the hash there is load bearing (especially since certificate checking is disabled in nixpkgs fetchurl if a hash is given, so the HTTPS doesn't protect you)

[XYZ] ww 🩶 @ww@xyzzy.link

@xerz @star didn't know everyone on fedi was a crc32 fan 🙃

cat-shaped cat @l@lea.pet

@star apparently windows still hashes passwords with MD4 :)

❤️2

flori_ava_star@[matrix]conf:~

@star@amazonawaws.com

@l no there it's fine. they need it for their backdoors

❤️3🩷1

flori_ava_star@[matrix]conf:~

@star@amazonawaws.com

???????????,,?????,!!,!,!,?.

@l my backdoor is hashed with md1 the way it is widely accessible

:neocat_googly_shocked@mk.absturztau.be:

1❤️1😭1🩷1

Emma, die Goth-oma @emma@gts.apicrim.es

@l @star wait wtf

cat-shaped cat @l@lea.pet

@emma @star backwards compat for smb 1 or some bullshit from what i know

Emma, die Goth-oma @emma@gts.apicrim.es

@l @star okay, well, that alreqday isn't an explanation I accept wtf are they smoking

We have had collision attacks for over 30 years now!

cat-shaped cat @l@lea.pet

@emma @star its fine :)
microsoft knows best :)

Emma, die Goth-oma @emma@gts.apicrim.es

@l @star microsoft knows best.

V-486233, b0rk3d c0rp0 dr0n3 @vrn@app.wafrn.net

@star

given the illusion of free choice, I'd pick Blake3 100% of the time.

https://github.com/BLAKE3-team/BLAKE3

Makuru

(📞 WAAF!) @makuru@catgirl.cloud

@star Not even for simple transmission error checking?

⛧-440729 [sophie raven] @sophie@catgirl.cloud

@star meanwhile, git commits:

❤️1

⛧-440729 [sophie raven] @sophie@catgirl.cloud

@star (but yeah anything that actually needs to be authenticated in any capacity, do not use)

Lambda Crime [The Möth Society] @16af93@wetdry.world

@sophie @star the humble sha256 git commit: [violently explodes because for some reason some tools dislike working with sha256 commits]

[4fbb09] @5225225@furry.engineer

@sophie @star hey, it uses sha1dc, which probably patched the collision method that we know about so far :) (and there's probably a lot less interest in breaking sha1dc since sha1 was already demonstrated broken, why'd you need to break it again?)

❤️1

flori_ava_star@[matrix]conf:~

@star@amazonawaws.com

@5225225 @sophie Well the interest would come from how widely used it is, non?

❤️1

[4fbb09] @5225225@furry.engineer

@star @sophie true, though in git the impact is... minimal

and primarily a case of "you can't use git commit hashes of possibly hostile repos as a way to refer to something that you've audited to be trustworthy"

collision attacks aren't Good but for the case of "i am trusted and emit a md5, you go and download a file and use it if it matches that md5" it's... not Strictly Speaking an issue (assuming your input is entirely not attacker controlled)?

you'd need a second preimage attack for that to be dangerous, which afaik not even md5 is broken in that way yet.

not that you should rely on that property, but if you want to go and make a "incredibly suspicious looking but not technically broken(yet) cryptosystem" then relying on the distinction between second preimage attacks and collisions seems like a fun way to do it.

Lyx @lypsilonx@lypsilonx.de

silly, lewd

@star @5225225 @sophie Hehe… „widely used“

❤️2

flori_ava_star@[matrix]conf:~

@star@amazonawaws.com

silly, lewd

@lypsilonx @sophie @5225225 are YOU widely used, Lyx? :3

❤️1

Lyx @lypsilonx@lypsilonx.de

silly, lewd

@star @sophie @5225225 No

Lambda Crime [The Möth Society] @16af93@wetdry.world

@star the humble git ecosystem:

❤️2

Lambda Crime [The Möth Society] @16af93@wetdry.world

@star (using sha256 - even tho formally implemented - tends to break stuff)

❤️2

zoe

@commanderred@chaos.social

@star may I speak to you about git?

Max @korenchkin@chaos.social

@star not even for demonstrating how broken they are?

❤️1

Alba 🌸

@mildsunrise@tech.lgbt

@star many national documents and passports have news for you

jackson&

@jacksonchen666@micro.jacksonchen666.com

@star stares at mediawiki

Lilith Ashley Nyx Arson

@root@possum.city

@star alright fine I'll use SHA-0 /j

❤️1

[ASTRL] Lyna @lynastral@catgirl.center

fr*nch joke

@star i use chat1 instead

icky yeen @niki@bark.lgbt

@star i heard from a trusted source that you can reliably use the length of a string to hash stuff... {

}

icky yeen @niki@bark.lgbt

@star that was alluding to this: news-web.php.net/php.internals/70691

quote: "Back when PHP had less than 100 functions and the
function hashing mechanism was strlen()." [...]

JCM @jcm@wafrn.jcm.re

@star

Yeah, fully agree, there's no reason to use anything past CRC32 :P