Introducing Authgasm: "Rails authentication done right".
binarylogic.com
binarylogic.com
SHA256 is slower than SHA1, which is slower than MD5. But in the grand scheme of things, SHA256 is fast. It's designed to be fast --- fast enough to run on a per-packet basis in secure network protocols.
Bcrypt is designed to be slow. So is "stretched" SHA256, which is simply SHA256 iterated thousands of times. Both these schemes are much harder to crack than simply hashing a password with SHA256.
The trouble I have here is that authentication is one part of your application you don't want to mess around with, but already someone here said "I'm going to try this just because the name is awesome". Forget the cosmetics. There's only one right answer to this problem.
However, I note that Coda Hale appears to have no interest in supporting his bcrypt-ruby gem for Windows. On his blog he says essentially that he doesn't use Windows and hence has no motivation to put any time into getting a pre-compiled bcrypt-ruby to work on Windows.
I can totally understand where he's coming from and as I do all my dev work on Debian Linux, it's not a problem for me personally. But my partner is a Windows user and it would be a problem for him.
The closest I have found to instructions on getting bcrypt-ruby compiled on Windows are from the bcrypt-ruby README:
You‘ll need a working compiler. (Win32 folks should use Cygwin or um, something else.)
So does this mean that my partner would then have to use the Cygwin environment for all his dev work from this point on?
And a random comment from Coda Hale's blog:
bcrypt seems to work with the Windows One-Click-Installer too. It was easy to compile it with MinGW (I just had to define the missing types u_int8_t, u_int16_t, and u_int32_t). All your tests pass. So you might consider to offer a precompiled version for us Windows users.
I am not a C programmer (getting more experience with a compiled language is on my todo list, just no time as yet) nor am I a security expert. My past experience with compiling stuff basically extends to typing "make" and "make install". I don't really understand what "I just had to define the missing types u_int8_t, u_int16_t, and u_int32_t" means.
So as much as I would like to try and incorporate bcrypt-ruby into my current Rails project, I can't really justify this added overhead of getting it to work on my partners Windows dev box with near zero support.
If anyone has more info, that would be great. But otherwise, I will - for now - be sticking with salted SHA256.
Who cares about the name. I wanted something unique that would get people's attention and return unique content when searching. I also thought it was funny. People need to lighten up a little bit if the name bothers them. It's not like I put the name "authgasm" in helpers that you have to use all over your application. It sits in your plugin dir, that's all.
Anyways, regarding encryption. You can encrypt the password any way you want. Use bcrypt-ruby if you want. It would take no time at all to set up. Checkout the acts_as_authentic documentation. You can provide your own "crypto provider" and have it do whatever you want.
Lastly, I don't use Sha256 straight up. I add in salt. So reverse lookups are out of the question. Your last option is brute force. There is no encryption / hashing algorithm that is safe from brute force. Granted some might slow it down more than others, all that is doing is setting up more hurdles to jump, not 100% secure.
But I really like bcrypt, I'll look into it and maybe switch it to the default crypto provider. But, I really think for anything that is not being used in the NSA, Sha256 + salt would work just fine.
The main issue I have with bcrypt-ruby as I said is that there is basically no Windows support for it that I can find. I know that most of us who develop with Ruby/Rails use a Mac or Linux/Unix but there are some, like my business partner, who still use Windows as their development environment. For a Rails plugin or Ruby gem, I think it would be really nice if at least some effort was made to keep it more easily cross-platform compatible for those of us who are not C coders.
But, I really think for anything that is not being used in the NSA, Sha256 + salt would work just fine.
As I said, I'm no security expert, but the reading I have done would lead me to agree with you. It's all about degrees of security with the trade-off always being more security means less convenience for someone, be it the developer or the end user. I've weighed the security/convenience issue for me at this point in time, and I'm sticking with salted SHA256.
That said, I don't so much care whether you store your passwords with a demonstrably inferior scheme like Authgasm's default or the 1-line-delta variant of it that resolves its biggest problem. It's fine to be ignorant about this stuff; it's not going to make you 1 extra dollar to do it right.
Just don't be militant about your ignorance.
I don't think I'm being militant about anything... Just carrying on what was a reasonably pleasant and honest conversation. Yes, part of my decision was to do with that my partner is a Window's user and I can't do anything about this at the moment. But the other part of the calculation that you overlook is that my current project is not one that requires over the top, secret squirrel encryption on stored passwords. This consideration came into it as well.
You sound bitter, man. Maybe you should take a nap and stop being so militant yourself.
Did the SHA eat your babies or something? :P
Seriously though, probably "over the top" was not a good choice of term. In a perfect world, I agree with the unspoken point you're trying to lead me into, which seems to be that all applications should never compromise on their security regardless of convenience or who their users are etc. But surely you must agree that in the real world this isn't always viable. I don't have unlimited time at my disposal.
If my current situation allowed it, I would use bcrypt based solely on the fact that OpenBSD recommend it.
However, my current situation does not allow this. Therefore, I have made a pragmatic decision that will enable me to continue moving towards my goal. I'm not going to apologize to you or anyone else for this.
I appreciate that you are passionate about what I assume is your chosen field of expertise, but perhaps in the future you could try talking to people more so than down to them. You might just find it yields friendlier results.
Otherwise, your users aren't simply giving you "shit passwords". They're giving you their bank password. Normal people do not make up 16 different passwords for all their web applications. If you take passwords from normal people on the Internet, and you ever get popular, you will come into possession of a large database of Bank of America credentials.
So yes, maybe you ought to consider being careful with them.
In any case, I accept that there should be some degree of responsibility taken by the developer to try and protect their users information, lest we get the whole Reddit situation.
But as someone once said, you can't make the Internet idiot proof.
Regardless, you can't realistically force your users to not have a "shit password policy". What you can do, however, is make sure that your app is probably not the point of failure for their password.
You can't make the internet idiot-proof, but you can make your apps reasonably malicious-person proof, and it's viewed by some (me) to be close to a moral responsibility of yours to do so - you either already know about, or are probably easily able to learn about "good" security policy, and ways to store users passwords and do validation while minimizing the risk of a malicious person acquiring your users passwords.
That is not to say that writing your own authentication / password storage scheme is something that you just shouldn't ever do - it can certainly be educational and entertaining - but 9 times out of 10 what you write probably shouldn't be used in production.
This whole conversation is based on the above users belief that unless you are using bcrypt to encrypt the passwords in your database, you are doing a moral disservice to your users. I and at least one other guy have chosen for varying reasons to use salted SHA256.
If anyone can prove to me that salted SHA256 is easily crackable, not by NSA super computers, but by some guy who just might want to steal the database of my little application and crack my users passwords, then I will strongly reconsider my choice. This is not a dick size competition, I'm not a security expert, I'm not challenging anyone to prove me wrong, because I don't think I have the first hand knowledge to know if I'm right or not. But all the material I have found so far tells me that salted SHA256 is still viable encryption for my purposes.
If it's not, then I am genuinely interested to see the conclusive proof.
I wasn't trying to come of as critical of you, although I can see how my post could be read that way - I was just making the point that homegrown stuff in this area has a notorious reputation for being unreliable.
I would argue that if using bcrypt makes your passwords harder to crack than using SHA256, there isn't much of a reason to keep using SHA256 - salted or not - regardless of how hard SHA256 is to crack.
And yes, 90% of the time, authentication / password storage schemes exist for whatever framework or platform that you're using that have been thoroughly tested by people who probably know their stuff better than you or I - and, from a security standpoint, it's probably better to use their scheme as opposed to writing one yourself.
However, as far as I can tell, none of the ready made authentication/password storage schemes for Rails use bcrypt in a cross-platform way, if they did I'd switch over. Where does that leave us? You can take that as a rhetorical question if you like.
If you don't understand the problem you're trying to solve with this plugin, why are you promoting it?
I actually looked into bycrypt and it's great. But it would be silly of me to use a linux only encryption solution in a library available to the ruby public.
http://www.binarylogic.com/2008/11/22/storing-nuclear-launch...
Authgasm.