On 8/22/07, Kim Goldenberg <[EMAIL PROTECTED]> wrote:

> I don't just blindly remove security functions just because it "gets in
> the way". Ive even set up ssh keys with non-null passphrases as well as
> ssh-agent, to verify it's me and not someone who scarfed up my key
> without my knowledge.

:soapbox.
This is not a matter of getting in the way. What does get in the way
is a root password that is known by some people and can be used beyond
their original need to know. If you have 100 Linux virtual machines
used by various people, it just does not work well to invent 100 good
passwords every month to give each team the proper access.

Acceptance by others gets very low when they cannot have a root
password, but you can... and they will come up with a manager to
approve that they put the root password in some silly automated ftp
that copies data from one system to the other...   Not having a root
password is the best way to get out of that.

We've used this and it really works. We did have a server virtual
machine play SCIF (with logging and auditing and access control) for
when no ssh login was possible, or for automation things.

The good thing about cryptic keys is that you separate authentication
and access control, which we believe is a good thing to do. It
provides granularity and ease of use. When you already have your
workstation protected well enough, ssh-agent makes it very easy indeed
(and secure because people don't see you type in a password).
Even if you have to type your passphrase each time, that's probably
more secure against people reading it over your shoulder (because it's
the same for all systems and you can probably type it very fast). Way
better than having to look up the root password for server #86 when
someone is watching you...

And non-encrypted private keys (null passphrase) are evil. Except for
cold bodies (i.e. not warm bodies, so machines or automated processes.
And obviously you make sure that such a key only gives access to what
that process must do..

The authorized_keys file for root on that server gives full access
control. And it does auditing too. You can also use this for db2inst1
or whatever functional accounts you have. And it does not have to be
the same list of users who have access. If you want to go fancy, you
move the authorized_keys into LDAP and get the ability to build groups
and update access without messing with individual systems.

PS I believe we were eventually forced to have a root password because
corporate standards dictate that you change it every nn days, and if
you don't have one you cannot check that it expires every nn days :-(
 So I think we eventually set random passwords that nobody knew.

Rob

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to