[cryptography] Constitutional Showdown Voided as Feds Decrypt Laptop

Steven Bellovin smb at cs.columbia.edu
Thu Mar 1 20:58:38 EST 2012


On Mar 1, 2012, at 8:18 32PM, Jeffrey I. Schiller wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> On 03/01/2012 06:09 PM, Nico Williams wrote:
>> I let mailman generate passwords.  And I never use them, much less
>> re-use them.  Well, I do use them when I need to change e-mail
>> addresses, which happens very rarely, and then I start by asking
>> mailman to send my my passwords because I don't remember them -- I've
>> done this like once in the past decade.
> 
> Perhaps mailman should be changed to require you to use its generated
> passwords, or better yet, to only generate a password when you ask it
> to send you your password, and then invalidate it after a few days. So
> it isn't really a password but a "thunk" of limited value.
> 
> In this fashion we can be more assured that people aren't re-using
> passwords with mailman.
> 
> Because... you and I may know better... the manager at the bank where
> are money is stored (or the doctors office where our medical records
> are located) may not know better...   ;-)

(typo corrected above.)

Not a bad idea, though I'm not certain it's worth it.  Fortunately,
since the default is for it to auto-generate its passwords, they're
not likely to be used elsewhere.  I'd wager long odds that most people
never even use that password.  (And the bank or the doctor's office?
They're not using mailman, because it would take a sysadmin to install
it for them...)

In an ideal world, perhaps this isn't necessary.  Mailman would somehow
learn everyone's email public key, to send passwords encrypted.  
Alternatively, it could somehow learn your web public key -- an in
particular, the one you use for this mailing list -- and use it to
verify the client-side cert you use to log in to this particular
mailing list.  (It can't be just any cert you have, since of course you
have many of them to avoid being tracked.)

Better yet, it could do a remote read on your /dev/brain and *know*
when you wanted to log in, weren't under duress, etc.  I regard that
as about as likely as the public key alternatives, at least if we're
sticking to the real world.  

Turning back to your specific suggestion: that sets the security of
your mailman account to the security of your email account.  Of course,
that's what the current scheme does.  The secret is valid for longer,
but I'm not convinced that that matters all that much.


		--Steve Bellovin, https://www.cs.columbia.edu/~smb








More information about the cryptography mailing list