Linus on Git and SHA-1

https://plus.google.com/+LinusTorvalds/posts/7tp2gYWQugL

by dankohn1 • 9 years ago

550 170 9 years ago

plus.google.com

claar• 9 years ago

Also see discussion of Linus's earlier comments at https://news.ycombinator.com/item?id=13719368

bascule13• 9 years ago

Linus's transition plan seems to involve truncating SHA-256 to 160-bits. This is bad for several reasons:

- Truncating to 160-bits still has a birthday bound at 80-bits. That would still require a lot more brute force than the 2^63 computations involved to find this collision, but it is much weaker than is generally considered secure

- Post-quantum, this means there will only be 80-bits of preimage resistance

(Also: if he's going to truncate a hash, he use SHA-512, which will be faster on 64-bit platforms)

Do either of these weak security levels impact Git?

Preimage resistance does matter if we're worried about attackers reversing commit hashes back into their contents. Linus doesn't seem to care about this one, but I think he should.

Collision resistance absolutely matters for the commit signing case, and once again Linus is downplaying this. He starts off talking about how they're not doing that, then halfway through adding a "oh wait but some people do that", then trying to downplay it again by talking about how an attacker would need to influence the original commit.

Of course, this happens all the time: it's called a pull request. Linus insists that prior proper source code review will prevent an attacker who sends you a malicious pull request from being able to pull off a chosen prefix collision. I have doubts about that, especially in any repos containing binary blobs (and especially if those binary blobs are executables)

Linus just doesn't take this stuff seriously. I really wish he would, though.

tytso3• 9 years ago

That's not the plan. That was an idea that was thrown out if this was an emergency (it's handling different length hashes, and doing so that we don't have to force a flag day conversion which is hard), but once people realized that in fact, the sky was not following, the plan which Linus outlined in his G+ post was devised --- which does not involve truncating a 256-bit hash.

bascule13• 9 years ago

Can you please link me to "the plan" then? I have been trying to follow some of the ML discussion and that was the last plan I saw him put forth, e.g.:

https://marc.info/?l=git&m=148787047422954

greglindahl• 9 years ago

Do you have any insight into the sha-256 vs sha-512 choice?

eslaught• 9 years ago

Can you link the new plan? I didn't see it described in his G+ post.

ksherlock• 9 years ago

On the mailing list he suggested displaying 160 bits for backwards compatibility with existing tools. The full 256 bits would be used internally (and displayed externally with the proper --flag)

diminish• 9 years ago

additionally cosmetics but instead of hex string could they not use a-zA-Z0-9 in the visual output to the user to make the git command line text output shorter?

0efaa. -> 2AdC..

paulddraper2• 9 years ago

especially in any repos containing binary blobs

Yeah, that part is the real flaw in the argument.

hermitdev3• 9 years ago

Although, I generally defer to someone such as Linus in having far more domain knowledge such as this, but I'm concerned willingness to just drop bits of the hash like this. I mean 20-30 years ago, I could quasi understand for the sake of performance, but are we really so concerned about performance of a few clock cycles vs opening ourselves to a known vulnerable attack?

hdhzy• 9 years ago

I think the truncation is considered because a lot of software assumes git commit id has 160 bit length. It's not because of performance.

hueving4• 9 years ago

What is this known vulnerable attack you are talking about? The birthday paradox applies to every hashing function.

the_mitsuhiko• 9 years ago

Linus's transition plan seems to involve truncating SHA-256 to 160-bits.

This is not the plan. This is the backwards compatible system for tools that can only deal with 40 characters.

chainsaw10• 9 years ago

reversing commit hashes back into their contents

Somewhat off topic, but is this actually possible?

Given hashing is inherently lossy, I'm inclined to assume it's not possible for anything must longer than a password, but commits are text, which I suppose is low entropy per character, so I don't know.

bascule13• 9 years ago

Yes this is possible. It's called a preimage attack:

https://en.wikipedia.org/wiki/Preimage_attack

It's computationally infeasible against most cryptographically secure hash functions. However, Grover's algorithm results in a sqrt(keyspace) reduction in security levels (effectively halving bit-sized security levels):

https://en.wikipedia.org/wiki/Grover's_algorithm

pshc2• 9 years ago

It doesn't sound possible for all but the most trivial diffs. An attacker would have to guess not only the exact commit text, but also its timestamp. And the only attack scenario I can conceive of seems pretty silly. Maybe I'm missing something...

Alice clones an open source git repo, commits one secret change where she edits a config file's default password to her own secret password (a bad practice), and then publishes the new hash in public for some reason (build info?). Mallory would have to (a) know that exactly this happened, (b) guess the commit message, (c) guess the commit's timestamp to the second (or within a few seconds), and (d) preimage-attack her password.

And the preimage attack must pierce git's Merkle tree, which sounds downright impossible. (Unless Mallory is just bruteforcing, in which case a strong password is enough.)

hueving4• 9 years ago

No, you're correct that for large inputs there are obviously going to be many that lead to the same hash (see pigeonhole principle).

usrusr2• 9 years ago

Most commits are not just text, they are source code. If you had a feasible way to enumerate all the byte sequences that collapse into a given hash value, you might find the subset that is syntactically correct code to be very low. Except if it's Perl, of course.

And the likelihood of all the characters aligning in a way a compiler might find acceptable gets lower with every increase in length of your collisions, so it would be extremely unlikely that the shortest nontrivial match (for both the hash and for code sanity) would not be the right one. The code constraint certainly would not make it easier to find the collision in the first place, but it would give you great confidence in the result of you did.

hueving4• 9 years ago

Preimage resistance does matter

Is there a preimage weakness though? I thought this attack only reduced collision resistance.

OJFord• 9 years ago

GP said that in reference to quantum attacks.

mappu• 9 years ago

(Also: if he's going to truncate a hash, he use SHA-512, which will be faster on 64-bit platforms)

BLAKE2 is faster still! It's also at least as secure as SHA-3, and produces any choice of output size up to 512bit.

stouset• 9 years ago

I'm a fan of blake2, but SHA-2 is the conservative choice here given the sheer amount of cryptanalysis it's been subjected to.

gkya• 9 years ago

If a repo contains binary blobs, especially executables, well, that's very bad practices right there. Also, how can somebody else modify a binary in a meaningful way and send a patch to it? How can you review a patch to a binary file before applying? I'd say that any sane project, especially if open source, would not include binaries (maybe apart from images), and even if it did, would not accept patches to them (if the members of a project were talking about changing, say, an icon, valid images would be exchanged on a mailing list/issue tracker, nobody would bother making, sending and applying binary diffs; then somebody w/ commit bit just commit it).

Let me put it more simply: if you're accepting patches for binary files in your repos you don't care about security at all. Maybe unless if you know how to decode machine code/JPEG manually.

Also, the proposition was to use the full hash internally, and truncate the representation. Nobody other than git itself uses full hashes anyways.

wongarsu2• 9 years ago

Executables in git are bad practise, but not all that uncommon. Images in git are the norm, and if somebody comes in and creates a pull request with an improved version of the existing images (better compression, better adapted for color blind people, fixing whitespace issues etc.) that's pretty unsuspicious and likely to succeed (and I've seen it multiple times).

Dylan168074• 9 years ago

What level is considered secure, then? The numbers for O(time) and O(space) should be many orders of magnitude apart to represent the relative costs.

simplehuman3• 9 years ago

Linus just doesn't take this stuff seriously. I really wish he would, though.

Can't downvote this enough. This is plain FUD. Did you even read the complete thread on the git mailing list? This was just one proposal by him.

AgentME4• 9 years ago

If he took this stuff seriously, he wouldn't have waited 12 years since SHA-1 was broken to even start considering any changes.

infinity03• 9 years ago

"A hash that is used for security is basically a statement of trust [..] In contrast, in a project like git, the hash isn't used for "trust". I don't pull on peoples trees because they have a hash of a4d442663580. Our trust is in people, and then we end up having lots of technology measures in place to secure the actual data."

This is horseshit, and Linus should not be saying these hugely misleading statements about security principles.

The point of a hash is to remove the need for trust between the trusted person who tells you the hash and the infrastructure you get the actual data that was hashed, from (edit: and, between you and the latter).

In other words, once you get a good non-colliding hash from a trusted person, then you don't need to worry about malicious infrastructure sending you bad data claiming to be the source of that hash.

Linus trusting Tytso to sign the commit object that references the SHA-1 of the tree object, says nothing about whether the infrastructure served him the tree object correctly. Sure, he might also trust the infrastructure providers, but when he says "trusts people" it does not sound like that is what he means. And even if he trusts the infrastructure providers, with a good hash HE DOESN'T HAVE TO.

The "trust" wording is serious horseshit.

(edit: there is also the case of people downloading "linux" from random git repos in the future. Right now if you GPG-sign a commit or tag, it has SHA-1 references to the tree object underneath it. Once SHA-1 is more broken it basically means you shouldn't trust random git repos across the internet to give you good content, even if it's "signed by Linus".)

bascule13• 9 years ago

Yes, exactly.

If Linus truly doesn't care about security, then git could use any error correcting code that produces a uniform distribution of tags, such as CRC64. The size of the tag only affects the number of objects we'd expect to be able to commit before we see a collision: over 4 billion in the case of CRC64.

Linux mistakenly claims using a cryptographic hash function helps avoid non-malicious collisions, but this is not the case.

Where the choice of a cryptographic hash function matters is specifically if we expect an attacker to be trying to collide tags. CRC64 is a linear function of the input data and therefore fails miserably at preventing attackers from colliding tags, but still produces a uniform distribution of tags for non-malicious inputs.

git seems to be in the odd place where Linus argues he's using a cryptographic hash function but not for security purposes.

ploxiln4• 9 years ago

You're saying Linus's statements are "hugely misleading", but it's just that you wish git were designed to be used differently. So, your argument is "horseshit".

Linus could have designed a cryptographically perfect system such that he could pull Tytso's signed commit from anywhere on the internet - but he didn't

Linus used sha1 as a useful tool for an effective DVCS with an initially simple implementation.

He still depends on the security of the kernel.org servers, his work computers, and the top submaintainer's work computers. His git trees and all the submaintainers he pulls from are hosted on kernel.org servers. Security of the kernel.org servers is taken very seriously, especially since the well-known break-in a few years ago. Now even two-factor auth is involved in all git pushes to kernel.org servers. Sub-maintainers make pull-requests by branch name - "please pull branch for-linus ...". More peripheral contributors submit their work via patches on LKML, no commit hashes involved.

Finally, no well-known SCM previous to git was based on perfect cryptographic proof of source history, or anything like that. It wasn't a big issue, and it's not the problem git focused on solving. Before git, we all used CVS, SVN, tarballs, and patches. And a significant portion of developers did not use any VCS at all. How could any of us have trusted any source code before 2005?! Somehow we did, though...

azernik6• 9 years ago

This should really be a top-level comment.

amluto• 9 years ago

- Post-quantum, this means there will only be 80-bits of preimage resistance

Post-quantum, it's only ~2^53 work to find a collision. IMO that's worse.

ploxiln4• 9 years ago

If anyone is really interested in more assurance of git commit contents, there's "git-evtag", which does a sha-512 hash over the full contents of a commit, including all trees and blob contents.

https://github.com/cgwalters/git-evtag

debatem16• 9 years ago

I'm mystified as to why this is even a discussion.

SHA1 is busted. That impacts some git users. The fix is not invasive. Fix the bug. Make the transition. Move on.

Super unprofessional.

AsyncAwait• 9 years ago

It's not that simple. Git is a widely used software integrated into many places, so keeping some backwards compatibility is important.

Just going ahead and start breaking things would really be unprofessional.

debatem16• 9 years ago

I'm not saying he should go out and break things now. I'm saying that the way forward he outlines could have been followed years ago, as essentially the rest of the industry did. Instead he waited until his user base was totally panicked to address their concerns.

In my view, that's not the right-- or even a rational-- way to do things.

snakeanus• 9 years ago

They did break backwards compatibility with git v2.0 but sadly they did not bother to change the hash function.

spatulon• 9 years ago

"Unprofessional" is an odd choice of word. I cannot see what you mean.

He explains why he wrote the post: "since the SHA1 collision attack was so prominently in the news." I think a lot of people on Hacker News and elsewhere are interested in both the SHA1 collision and how it affects git, even if it doesn't impact them.

Still, even if this were some obscure bug that nobody cared about, what would be unprofessional about explaining his plan to fix it? It's an open-source project, after all.

If anything, I'd say it's quite professional that he's calmly explained why there's no need to panic, that they have everything under control, and that the path forward should be smooth.

debatem16• 9 years ago

The issue is that this could have been fixed years ago. All the information needed to move forward today was also available the last time this got brought up. Fixing it then would have better for git's users (because they wouldn't have freaked out over this) and better for git (because it never would have even come up in the SHA1 discussion) and better for Linus (because he wouldn't have had to spend an hour writing a blog post about this). So why not do it then?

It's frustrating, because Linus clearly isn't stupid. And yet sometimes he does stuff like this where I can't help but go "how did you not see the unforced error you were making?".

cwbrandsma• 9 years ago

What is the realistic worst case situation here? From what I understand, git uses SHA1 as a way of generating an id for a file, not for security. So two files might match up when they shouldn't? Is that it or is there more to this?

Or is this a proverbial "since SHA1 is don't work over there, then it shouldn't be used anywhere under any circumstance" attack?

debatem16• 9 years ago

As Linus touches on in the article, SHA1 is used for signing in git. This is clearly a security function, and should not have depended on SHA1 for at least the last several years.

Additionally, the defense offered against substituting binary blobs essentially comes down to "well, the kernel doesn't do that". Respectfully, other projects do. Those users' concerns were not taken seriously until lots of unrelated users freaked out about SHA1 for bad reasons.

Regarding the worst case scenario, yeah, substituting one binary blob for a different one (say, in Google's AOSP git repo) would be the worst. Which isn't sky-is-falling bad, but would still be pretty ugly.

vacri• 9 years ago

Perhaps you should re-read the last paragraph of the article.

debatem16• 9 years ago

Can you explain why? I went back and reread it and it said the same thing I thought it did: that the fix wasn't a big deal and they were moving forward on it.

I agree it isn't a hairy fix, but think that it would have been better for everybody if it had been made before a SHA1 collision was found, as has been repeatedly proposed to Linus in the past. This would have avoided the entire discussion about SHA1 and git, and all the angst that went with it.

dboreham• 9 years ago

Um what? Software written in the past 20 years has a baked-in assumption that the length of some ID can't change?

theseoafs• 9 years ago

Have there been writings on what exactly git's migration strategy to a new hash function will be? Apparently they have a seamless transition designed that won't require anyone to update their repositories, which seems like a pretty crazy promise in the absence of details.

keeperofdakeys• 9 years ago

In git the SHA-1 hash is simply an identifier for an object - it's used in the filename, but not stored in the object. And when a commit or tree object references others, it's just a name that can be looked up in the database. So a commit object hashed with SHA-256 can easily reference a previous commit that was hashed with SHA-1.

During the switch, a bit of deduplication may be lost. But the only interesting issue I can see is how git fsck will tell which hash an object was created with when verifying the hash (maybe with length?).

iamgopal• 9 years ago

Git update repo kind of command may be ??

runeks3• 9 years ago

One thing SHA-256 has going for it is that millions can be made from finding pre-image weaknesses in it, because it's used in Bitcoin mining. If you could "figure out" SHA-256, and use it to take over Bitcoin mining, you'd make $2M the first 24 hours, at current rates. And if you play it wise, it could take a long time before anyone figure out what's going on.

With regards to market price for a successful attack, I don't think any hash function stands close to SHA-256. And for that reason I think it would be the right choice.

wolf550e2• 9 years ago

https://twitter.com/veorq/status/834872988445065218

runeks3• 9 years ago

I think the claim is interesting, and I certainly wouldn't reject it. But just relaying someone else's claim, without any substantiating argument, I find quite uninteresting.

kzrdude3• 9 years ago

SHA-512/256 is faster and has the same output length

colin_fraizer3• 9 years ago

This, btw, is why we have e-cigarette bans. The fact that the generally high-IQ, paid-to-think-about-subtle-categorization community of software developers needs to be inoculated against the "I Heard SHA-1 Was Bad Now" meme, should serve as a reminder for why most things should not be managed by democracy.

(Yeah, I know this will be read as a plea for monarchy and downvoted. It simply proves my point: people are WAY too subject to errors in the classes (1) "I hate him because he said something 'bad' about something 'good'." and (2) "I hate him because he said something 'good' about something The Tribe now knows is 'bad.')

grzm4• 9 years ago

Save yourself some downvotes and remove the mention that you expect them.

colin_fraizer3• 9 years ago

Save myself from people proving my point? Why?

butwhynotmore• 9 years ago

In the specific case of cryptography where it's unknown how bulletproof the algorithm will be why not use multiple hash functions? Perhaps using the top 10 best hash functions of the day. That way you're not putting all your eggs in one basket and if nefarious collisions are able to be created in the future you still have the other hash functions to both "trust" and double check against. It's even more unlikely that nefarious collisions will be able to be constructed that collide all the other hash functions as well. You could just append the hashes to each other or put them in a hash table or something. Maybe my computer science is not up to snuff but it seems like this would provide more resiliency against future and non-public mathematical breakthroughs as well as increased computing power such as quantum computing. Yes, it would take a little longer to compute all the hashes in day to day use, but with the benefit of a more robust system both now and in the future.

hannob3• 9 years ago

One thing that I think is worth mentioning: This was completely avoidable. Git isn't that old, it wasn't taken by surprise by the SHA1 attacks.

The first paper from Wang et al, which should've put SHA1 to rest, was published in 2004, the year before the first ever Git version was released. It could have been easy: Just take a secure hash from the beginning.

simias• 9 years ago

While this post sounds very reasonable to me there's one point that I really don't get: why does he keep saying that git commit hashes have nothing to do with security?

If he believes that, why does git allow signing tags and commits and why does Linus himself sign kernel release tags? Isn't that the very definition of "using a hash for security"?

jmount• 9 years ago

Probably isn'y the sky falling. But if knowing the length fixed all hash function issues then cryptographic hashes would just use a some more bits for length.

maxander• 9 years ago

I don't really get the threat model here. If an attacker is pushing commits into your repository, you're long since toast on all possible security fronts, right? Is there anything nefarious they could accomplish through hash collisions that couldn't be done simply by editing commit history?

azernik6• 9 years ago

If they edit the commit history and you're using a secure hash algorithm, then the hash of the current commit will change and no longer match the signed tag your trusted maintainer sent you.

leoh• 9 years ago

Not really. From Linus — I think the most important point that has not been discussed extensively:

But if you use git for source control like in the kernel, the stuff you really care about is source code, which is very much a transparent medium. If somebody inserts random odd generated crud in the middle of your source code, you will absolutely notice.

yeukhon• 9 years ago

Well, open source software like Linux kernel has many obvious bugs for years go without notice. But his point is still extremely fair. There may be code no one really actively updating, but to even able to add random odd crud in the source code is really really really really hard. The effort is not worth it for any reasonable attackers. One would rather exploit an existing zero day or just infect network and wait to activate the infection. If someone modifies the local source code you just git cloned (or from a tarball), well you have something to worry about - your network.

stinos• 9 years ago

If somebody inserts random odd generated crud in the middle of your source code, you will absolutely notice

Unless I'm misunderstanding how this would work in practice (I assume commits being added under my name or existing commits being modified?), I'm abolutely not buying that though. Way too general, no? I would likely notice something like that, but I know enough colleagues who are just as likely to not notice. Mostly because they don't fully grasp git so they're just like 'hmm, thing says I'm behind, ha I know the fix is pull, yes I'm a git wizard' and just pull then continue working without even checking what got changed.

btbuilder• 9 years ago

How about very subtle in with the noise. Like this attempt to backdoor Linux in 2003:

https://lwn.net/Articles/57135/

Hard to get a collision though I'm guessing.

posterboy4• 9 years ago

If somebody inserts random odd generated crud in the middle of your source code, you will absolutely notice.

Just as they noticed the intrusion on kernel.org servers after how long?

edit:

If an attacker is pushing commits into your repository, you're long since toast on all possible security fronts, right?

Sure, that's always true, in the worst case. Perfect security doesn't exist. But does it need to be an attacker? Couldn't it just happen that an important bugfix isn't recorded because the hash collides?

posterboy4• 9 years ago

random odd generated crud in the middle of your source code, you will absolutely notice

like non-printing characters in comments?

Or, you know, random odd generated crud https://github.com/torvalds/linux/tree/master/firmware/radeo...

hackuser• 9 years ago

Related, from Mozilla:

* The end of SHA-1 on the Public Web

https://blog.mozilla.org/security/2017/02/23/the-end-of-sha-...

As announced last fall, we’ve been disabling SHA-1 for increasing numbers of Firefox users since the release of Firefox 51 using a gradual phase-in technique. Tomorrow [Feb 24th], this deprecation policy will reach all Firefox users. It is enabled by default in Firefox 52.

azernik6• 9 years ago

To be fair (although I've been commenting angrily about git's continued use of SHA-1 elsewhere) it's a lot easier for a browser to change hash algorithms than for git.

joatmon-snoo4• 9 years ago

The actual mailing list discussion thread can be found here, and is infinitely more informative than any of the bull being spouted in this thread: http://public-inbox.org/git/20170226004657.zowlojdzqrrcalsm@...

azernik6• 9 years ago

My takeaway from all that - the git project is internally taking steps to mitigate vulnerabilities from this particular attack (making it harder to insert the arbitrary binary data necessary into git metadata), but a) is just throwing up their hands at the problem of projects that store binary blobs like image data in their repos, and b) is not taking this as a signal that more serious sha-1 attacks are on the horizon and they should speed up their hash-replacement efforts.

This latter leads into the problems with Linus' positions in particular. In that thread, does not take seriously the threats that this poses to the broader git userbase, because he only seems to care about the kernel use-case: trusted hosting infrastructure at kernel.org (itself an iffy assumption, given previous hacks and the use of mirrors), and the exclusive storage of human-readable text in the repo which makes binary garbage harder to sneak in. These do not apply to most users of git. His rather extreme public position (paraphrased, "our security doesn't depend on SHA1") is even more troubling - it absolutely does depend on SHA1, this just isn't (yet) a strong enough attack to absolutely screw over the kernel. A stronger collision attack (eg a chosen-prefix as opposed to identical-prefix, or godforbid a pre-image attack) would absolutely invalidate the whole git security model.

Chyzwar• 9 years ago

SHA1 is only used for uniqueness. You still need to have write access to repo to perform attack. If you already have write access you do not need to make collision...

0xbadcafebee• 9 years ago

Git has a security model?

edit: rephrase: Git has no security model.

tytso3• 9 years ago

That's not all "the git project" is doing. Work to allow git to support an alternate crypto checksums is on-going. One of the key steps is to replace the hard-coded use of char[40] with a "struct object_id" object, and it is about 40% done[1].

[1] http://public-inbox.org/git/20170217214513.giua5ksuiqqs2laj@...

Like many volunteer projects, the work would go faster if more people helped. What "the git project" considers important works the same way as "how much is bitcoin worth", or "how much is gold worth". People who think it's important can show the value by putting their effort where they think it is most important. Or, of course, people can snipe from the side-lines, and make themselves feel all self-important.

There has actually been a lot of thought about how to do a graceful transition period, but to also on how switch from a soft cutover to a hard cutover (breaking backwards compatibility and requiring people to upgrade their clients) to allow time for developers to upgrade at a reasonable rate, where reasonable can be defined on a per-git repository basis. So if you want to force a flag day as soon as the code is available, and not wait for stability testing, etc., those people who value security uber ales, and who don't want to rely on trusting kernel.org, etc., can do so.

I will note, though, that most people are doing blind pulls, or worse, blind merges, from developers outside of their immediate circle of trust _all_ _the_ _time_. Heck, people will cut and paste command lines from web pages of the form "curl http://alfred.e.numman/what/me/worry | bash" into root shells all the time! So if you are not auditing every line of code before a git pull, the fact that git is using SHA-1 is the least of your worries.

Personally, if I were a nation state planning on trying to insert the equivalent of a DUAL-EC backdoor into open source software, I'd do that by spending a person year or ten getting a collection of developers to be trusted contributors to some key open source project, like Docker, or Python, or even yes, the Linux kernel or git, and then "accidentally on purpose" introducing a buffer-overrun or some other zero day into said OSS code base. Or heck, just simply invest in finding more zero days that people have inserted into their code just because they're not careful!

So, sure, "we" should upgrade git to be able to support multiple crypto hash algorithms, and there is work going into doing this. But at the same time, it's important to keep a sense of perspective on all of this --- unless, of course, your goal is to make yourself seem important by exuding a sense of self-righteousness.

joatmon-snoo4• 9 years ago

a) is just throwing up their hands at the problem of projects that store binary blobs like image data in their repos, and b) is not taking this as a signal that more serious sha-1 attacks are on the horizon and they should speed up their hash-replacement efforts.

As @tytso points out, there is ongoing work to replace the use of SHA-1. Also, yes, SHAttered raises serious concerns about the safety of SHA-1, but that doesn't mean everyone has to immediately work on the switch (and honestly, it's questionable how much it would help - throwing more developers at a problem doesn't necessarily help). Should Stefan Baller drop the work he's doing on submodules because SHA-1 is more imminent now?

It's also worth noting Linus isn't really a core Git dev any more, he just submits patches occasionally. Junio Hamano is the primary maintainer.

yuhong2• 9 years ago

I do wonder how many outside of crypto circles know about SHA-2 circa 2004.

frik• 9 years ago

Can someone correct me. SVN/Subversion and GIT are affected by SHA-1 problem. SVN uses SHA-1 internally, but exposes only a numeric int as revision. GIT uses SHA-1 internally and as revision. So if someone commit a modified PDF that collides he can run havoc on both SVN and GIT at the moment. It seems easier to fix the issue in SVN than GIT.

hannob3• 9 years ago

It's a somewhat different issue.

Git can probably not be havoced by committing two colliding files (and doing so would require doing another chosen prefix attach with a git blob header). But git looses cryptographic integrity promises due to this attack (aka: you can have different source trees with different histories leading to the same top commit hash). svn never had any cryptographic integrity to begin with.

kzrdude3• 9 years ago

Sure, here's a correction: There has been no evidence of havoc or repo corruption in git.

It appears that to even try to attack git you'd need to spend the same amount of work again ($110K or what it was) to create a new collision with the right git object headers.

luckydude• 9 years ago

Linus is a little behind the times with this comment:

``Other SCM's have used things like CRC's for error detection, although honestly the most common error handling method in most SCM's tends to be "tough luck, maybe your data is there, maybe it isn't, I don't care".''

BitKeeper has an error detection (CRC per block) and error correction (XOR block at the end) system. Any single block loss is correctable. Block sizes vary with file size so large files have to lose a large amount of data to be non-correctable.

godzilla82• 9 years ago

Newbie question .. can some one please help me understand the attack scenario. if I, as the attacker, want to inject malicious code/binary into a git repo, then I need to write my malicious code/binary in such a way that the resultant hash collides with one of the commits (? Or the last one?) in the repo. Is this correct?

Linus on Git and SHA-1

Related Stories