[Cryptography] Cost of remembering a password

[Bear]

On Sat, 2014-08-16 at 11:38 +0000, Michael Kjörling wrote:

> Given how many people end up with passwords like "password1",
> "12345678" and so on, I think it isn't as much _passwords_ that need
> to be dealt with as password manager integration. All major web
> browsers for example have the ability to locally store passwords used
> (whether or not it's secure is a different matter and also depends a
> lot on the user's chosen master password/passphrase), but what is
> lacking is a _user friendly_, fully integrated, enabled by default
> means to automatically generate and store secure passwords, and with
> today's proliferation of different types of devices share passwords
> between e.g. a desktop computer, a smartphone and a tablet.

Whatever your password manager runs on, is a trusted system - 
ie, one whose compromise could absolutely destroy your security.  
And if it is a conventional system running software, then it 
is running something invisible and modifiable which I cannot 
fully inspect, ie, it is not trustworthy.  

We must never create trusted systems which are not trustworthy.  

A computer system which can reasonably be trusted not to be 
hiding a security breach cannot execute anything from RAM at 
all and has the only sofware it runs in firmware consisting 
of diodes (non-rewritable ROM).  Further, it uses the kind of 
embedded CPU that has no state, no cache, no firmware, and 
no bootup sequence.

Until you build one of those, I'm going to continue using an
iron box for a password manager.  Integrating a software 
password manager is a failure because it puts trusted software
on non-trustworthy hardware.

			Bear