lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  PHC 
Open Source and information security mailing list archives
 
Hash Suite for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 11 Sep 2014 13:46:22 -0700
From: Alex Elsayed <eternaleye@...il.com>
To: discussions@...sword-hashing.net
Subject: Re: Re: Re: Re: Re: Re: Re: Re: Re: Second factor (was A review per day - Schvrch)

Andy Lutomirski wrote:

> On Thu, Sep 11, 2014 at 12:27 PM, Alex Elsayed
> <eternaleye@...il.com> wrote:
>> Andy Lutomirski wrote:
>>
>>> On Thu, Sep 11, 2014 at 11:39 AM, Alex Elsayed
>>> <eternaleye@...il.com> wrote:
>>>> Andy Lutomirski wrote:
>>>>
>>>>> On Thu, Sep 11, 2014 at 11:26 AM, Alex Elsayed
>>>>> <eternaleye@...il.com> wrote:
>>>>>> Alex Elsayed wrote:
>>>>>>
>>>>>>> Mm, with the parallel PAKE we've latched P as a necessary part of
>>>>>>> the protocol, so _here_ your original idea of passing H(P) to the
>>>>>>> token, which it uses to encrypt its (internal) value, would not
>>>>>>> weaken the scheme in the case of a malicious token.
>>>>>>>
>>>>>>> Given token T holding secret X, user U holding password P, and
>>>>>>> server S:
>>>>>>>
>>>>>>> U -> T: H(P)
>>>>>>> T -> U: Y = E(k=X, H(P))
>>>>>>> T -> S: R_t = PAKE(X)
>>>>>>> U -> S: R_u = PAKE(Y)
>>>>>>> T -> U: R_t
>>>>>>> U: K = R_t ^ R_u
>>>>>>
>>>>>> After I posted the correction to this yesterday, I realized that
>>>>>> there's an even more optimal approach combining this with my original
>>>>>> scheme. It satisfies all of the properties you brought up, _and_ the
>>>>>> property I'd like that the server doesn't even need to know that a
>>>>>> token is in use at all (permitting a user to add a token to any
>>>>>> account they want to) - without requiring any local storage on the
>>>>>> user's machine.
>>>>>>
>>>>>> Let E( K, M ) be an encryption function with key K and message M
>>>>>> Let H( M ) be a hash of a message M
>>>>>> Let sizeof( H( A ) ) == sizeof( K in E( K, B ) ) for all A and B
>>>>>>
>>>>>> Given:
>>>>>>     Token T holding secret X
>>>>>>     User U holding password P and pin N
>>>>>>     Server S
>>>>>>
>>>>>> 1.) U and T establish a channel C by executing a PAKE over N
>>>>>> 2.) U sends H(P) to T over C
>>>>>
>>>>> This step loses the property that a malicious token doesn't weaken
>>>>> security over no token at all -- a malicious token can do an offline
>>>>> dictionary attack on the password.
>>>>
>>>> I'd argue that while it isn't _ideal_ (ideally, a malicious token
>>>> couldn't do that, I agree) it doesn't weaken security over no token at
>>>> all - because without a token (and thus X) the _server_ can perform a
>>>> dictionary attack on the verifier - so really, you wind up exactly
>>>> where you would be without the token as long as H is at least as secure
>>>> as the verifier generation algorithm.
>>>>
>>>
>>> Yeah, fair enough.  I guess that, if you're not willing to store a
>>> blob on your computer, then that's the best that you can do.
>>>
>>> On the other hand, for some token designs, you might have no choice --
>>> tokens with no storage will require you store a blob somewhere.
>>
>> True enough - my thinking has been along the lines of smartcards (or
>> GNUK), where the _cheap_ ones are storage only, and the rest are storage
>> + processing.
>>
>> Then again, a token with no storage at all can't really have much in the
>> way of a secret X - and if you _do_ store a (rather small - key-sized)
>> blob W on your computer, you can just replace H(P) with H( E( W, P ) ),
>> since H is only ever executed by U, thus preventing the token from
>> brute-forcing (since W is strong randomness &c)
> 
> I should have said ROM-only token as opposed to token w/o storage.
> Incidentally, Intel SGX looks a lot like a ROM-only token built in to
> a CPU :)
> 
>>
>> I just personally think the usability loss of storing a blob outweighs
>> the security loss of not doing so, especially since it needs to be
>> replicated to every machine that you want to log in from - making it (in
>> a sense) a poor hybrid between a software 'token' and a saved password.
> 
> I'm thinking of it like an ssh private key file with added two-factor
> goodness.

Well, my opinion on SSH private keys in files (as opposed to tokens) is 
"poor man's software token"...

It really is trying to fulfill "something you have", the passphrase for the 
key is "something you know" - but it's weakened as a "something you have" by 
being trivially cloneable.

It all comes down to the three proofs of identity - something you have, 
something you know, and something you are. Having _something_ is better than 
_nothing_ in each category, but you really don't benefit all that much by 
having more than one in a category.

You _do_ benefit in certain measures of security, but it starts costing more 
to get all of them to play nicely, and you start losing out on 
complexity/robustness grounds. To me, that's a huge problem, because 
Security Hole Zero is "the tools were a pain so I didn't use them."

It's the single most common security problem, whether it manifests as poor 
passwords, reused passwords, sticky notes with passwords on them, leaving 
two-factor authentication turned off because they don't want to type a 6-
digit code on login...

> If I find myself bored enough, I'll try to implement something in the
> Java card edition or whatever it's called and see if I can upload it
> to my YK NEO. *
> 
> * Old YK NEOs are unlocked and you can upload applets.  New ones are
> locked, but I think you can email Yubico and ask for the developer
> version.
> 
> --Andy


Powered by blists - more mailing lists