[Cryptography] Security weakness in iCloud keychain

[Ron Garret]

On May 8, 2018, at 12:01 PM, Kent Borg <kentborg at borg.org> wrote:

> On 05/08/2018 02:03 PM, Ron Garret wrote:
>> That’s it. We’re done here.
> 
> Let me see, how have I used passwords today?

Hopefully it was obvious that I was being a little intentionally glib.  I was going for a little (dark) ironic humor.  Nonetheless...

(I’m going to rearrange the order of your list a bit.)

>  - unlock my phone
>  - unlock my laptop
>  - sudo password on the account above

This falls under the purview of my scheme.  Instead of comparing your input password to a stored one, the device would see if the password you entered successfully decrypts your local keychain.

>  - password over ssh to a remote machine
>  - password over ssh to a shared account at work
>  - password over ssh to a shared account on an “appliance"

These are the canonical examples of where my scheme would work.  In fact, SSH already uses my scheme if you use ssh keys instead of passwords — which you should.  Password-based ssh auth is insecure.

>  - sudo password on the above account


This use case is a little trickier.  It would require sudo to be re-implemented to authenticate a local user whose credentials are not stored locally.  But it’s doable, and IMHO would be worth the effort.  The hard part is agreeing on a protocol.  The implementation would be straightforward, exactly the same as any other remote authentication (send nonce, check signature).

>  - wanting to click on a web link in an e-mail, but it will prompt me for a password and I never do things in that order, I manually log into that website (looking up the password because that one I do not memorize) to get a valid cookie and then I followed the link.

Sorry, I don’t understand this one.

> Six different passwords in the above.
> 
> With your tidy solution ("That's it. We're done here."), how many different passwords would I have typed so far today?

One.

rg