I tend to think of this stuff from the cloudcracker.com perspective, which is all about cheap jobs rather than millions spent. I rarely do brute force because, honestly, it's rarely necessary. But the single biggest factor in cloudcracker.com support of a hash format really is its efficiency in GPU space.

For instance, right now I'm implementing support for the modern SHA-512 crypt() variation. It doesn't translate well into GPU space at all, which will end up meaning that I can only offer dictionaries that are half a trillion words smaller than formats which are fast on GPUs.

So far the data I'm seeing indicates that differences of that scale in dictionary size really do make a difference on the success rate of the job. So for what it's worth, from that perspective, it is a factor.

I've yet to see a website that get bogged down with authentication request in normal operation, even if those request measure in the multiples of 10ms. The basic idea with all those "slow" password hashes is that authentication is a pretty rare request compared to all other request. Usually authentication requires a single request, at most a handfull. If you run into trouble with bcrypt/scrypt on your webservers, you're doing things wrong. If you have that many authentication requests, direct them to a dedicated server - you'll be in the league that has a large farm running anyways.

Granted, an attacker could use your slow hash for a DOS-attack, but most websites I've seen so far always had some sort of slow operation that was easily exploitable without authentication.

OK, but what if you're a web app that specializes in authentication (like, say, an Oath provider) or a database server where the attacker-facing app doesn't use connection pooling?

> but most websites I've seen so far always had some sort of slow operation that was easily exploitable without authentication

Yes, but those things are typically easy to optimize or temporarily disable in a hurry once they come under attack. Not so much with authentication.

> OK, but what if you're a web app that specializes in authentication (like, say, an Oath provider)

Then I'd hope that you make extra double sure that your user's passwords are secure. Let's call it the cost of business.

> or a database server where the attacker-facing app doesn't use connection pooling?

We're talking how to store user-passwords here, don't we? If your database server passwords get lost or cracked, change them. Use random-generated passwords, don't reuse them.

>> but most websites I've seen so far always had some sort of slow operation that was easily exploitable without authentication

> Yes, but those things are typically easy to optimize or temporarily disable in a hurry once they come under attack. Not so much with authentication.

That depends on the website. If your major content is available unauthenticated, then you might as well go and disable authentication. My major point was that you can't defend against a DOS attack by using a weaker password hash, the attacker will just throw more requests at you. In a DOS, the attacker does have the advantage of needing less computational ressources.

The discussion is about best practices and the relative merits of password hashing functions. So the baseline assumption is that the server-side password database isn't perfectly secure.

In practice, the user gets to choose the password and the website at best gets to veto it or accept it without knowing how many other places it's re-used. There aren't too many sites assigning randomly generated passwords right now, I wish there were more.

Yes, some DoS attackers may be able to throw more and more resources at you until you go down. But some don't and you don't have to make it easy for them by preemptively DoSing yourself with too much password hashing! Alternatively: for some fixed amount of attacker DoS resources, your system can support a certain amount of password hash cracking resistance. Cracking resistance is thus a tradeoff with DoS resistance. The root cause of this situation is the poor entropy present in many users' choice of passwords.

Turning off authentication is generally not an option if your site has any data worth securing. If it were, an attacker could bypass your access controls by simply DoSing you until you disabled authentication.

> The discussion is about best practices and the relative merits of password hashing functions. So the baseline assumption is that the server-side password database isn't perfectly secure.

That's right. So I don't see where database connection pooling is part of the issue and that's what my remark about random generated passwords was pointed at. You should use secure passwords to authenticate your app at the database. You should probably use md5 [1] or similar as password hashing scheme for the database credentials since this is where you really have a valid tradeoff between performance and security that allows any attacker to bog down your database. But this issue is not the scope of this discussion.

It's certainly a valid technical concern to not increase your password hashing work factor to a point where this it is a valid attack vector for DoS attacks, my point simply is that you can go quite far in terms of work factor until you reach that limit. Increasing the work-factor to a point where authentication takes 10ms will in many cases still leave other parts of your application more vulnerable.

And well, turning off authentication should imply denying access to data that requires authentication. This certainly is an option if the majority of your data is available for unauthenticated users. It certainly is not if you only store data worth securing.

[1] as postgres and mysql both do

Yep, I work close to an extremely high traffic OAuth endpoint. I think scrypt would probably "cost" a lot more in terms of actual servers required to operate it.

scrypt is tunable; you can make it use as much or as little CPU time as you want. For any particular amount of CPU time, scrypt will give you more security than bcrypt or PBKDF2 would give you for the same amount of CPU time.

Perhaps it's more accurate to measure scrypt in terms of cache misses rather than CPU time? This is a different resource on multicore servers with different performance implications (at least if we're down to counting 3 or 4 bits of security).

I thought the definition of "better" in this case is that it requires less work to get the same computational strength with scrypt. Are you saying that the memory locality issues that scrypt causes more than cancel out the computational win?

In this context, work is computational strength so what we're mainly concerned about is an attacker finding a way to do the task significantly more efficiently than the defender. E.g., if the attacker can evaluate the function with half the cost on his power bill relative to the defender, then that can be thought of as knocking off 1 bit of security off the top.

The primary advantage of Scrypt over the others is that it enters a completely pathological memory locality access pattern and stays there for almost the whole function. This works to neutralize the advantage of an attacker who has a custom CPU because he probably can't also develop a custom RAM subsystem to feed it with (at least not one that's many times more efficient than what the defender has in his server).

But if you've done any performance tuning on multithreaded code, you know that cache effects caused by memory access patterns very quickly begin to dominate as multiple cores and threads are added. Things that look great in single-threaded benchmarks almost never scale linearly and there's probably nothing that will scale worse on our shared-memory multiprocessors than Scrypt. It's a feature.

So the defender (say, a busy website with commodity multicore servers) with Scrypt is likely not going to be able to take as good an advantage of his hardware. He won't be able to crank up the work factor quite as high as he could with Bcrypt or PBKDF2.

This may represent an advantage to the attacker, who doesn't have the additional constraint of keeping the response time up on a busy webserver. This attacker's advantage is probably not significant by cryptographic standards (maybe 2 or 3 bits of security lost), but pathological multithreading could represent a big issue operationally.

I'm honestly not trying to cast FUD on Scrypt here, I think it's the best function. I'm just saying like everything else multithreaded you really need to benchmark it under real-world conditions.

scrypt's memory access pattern isn't particularly pathological; it's random, sure, but it reads large blocks.

The key issue isn't the design of the RAM subsystem but instead the design of the RAM subsystem -- in particular, making sure the attacker can't "cheat" by using a smaller circuit than the defender.