hi

All:

On 03/25/2016 11:23 PM, Anthony Carrico wrote:
>> I had a question about ssh passphrases. How many people use them for
>> ssh? Is it a Bad Idea not to use them? *If* I said I didn't use them,
>> would a collective gasp be heard across VAGUE land?

No worries, and thanks for asking! Being in VAGUE land does not
automatically imply strong knowledge of X. We are all experts in some
things, not all things. It is all about sharing what we know.

> Unfortunately, user applications typically aren't sandboxed in Linux
> distributions (Brian Waters, any info on this?), so they can use your
> key once your agent has the password, but at least if they snatch the
> key they'll need the password to use it elsewhere/elsewhen.

GNU/Linux does have "good" ASLR, not enough to stop all forms of Return
Oriented Programming, but quite a few are thwarted with ASLR. If you
want Mandatory Access Control, investigate SELinux, AppArmor, etc...

*I* strongly advise the use of passphrases, ssh-add, ssh-agent, and
friends. If you want a power-tool here, check out what monkeysphere
http://web.monkeysphere.info/ can do for you, some VERY cool actions.

> One application for a passwordless key is using ssh keys in an
> unattended system, for example a backup server. In that case only the
> backup service user should be able to see the private key on the server,
> and the client can increase security somewhat with from="..." in its
> authorized_keys file, see man sshd.

Agreed. I also encourage restricting what hosts are allowed for what
user. Also remember that hosts.{allow,deny} and fail2ban are useful
here, as well as monkeysphere.

In the case of monkeysphere, you can, for example, permit a new user
access to your ssh setup by key authentication using monkeysphere. A
very handy feature for remote/distributed setups--- a GPG key is
required of course...

>> If the weakness of the public/private key pair is the danger that
>> someone gets access to your file system and copies your private key, is
>> it recommended practice that the passphrase *not* be in a file somewhere?

If by "in a file" you mean in a plaintext file, yes, I avoid placing
plaintext keys on systems. FWIW: ssh keys encrypt the passphrase with
the key.

> If you put it in a file, use your PGP key to encrypt the file.
> 
>> Can anyone point to best practices for choosing ssh passphrases? I think
>> it needs to be long, easily memorable and highly entropic which all seem
>> contradictory.

There are many sources of advice on passwords, some good, some not so
good. With the addition of entropy in the form of numbers,
[upper::lower] letters, and punctuation, the XKCD method is a good
approach for improving password retention. https://xkcd.com/936/

That said, I strongly encourage the use of a password manager. <opinion>
*I* prefer one written in a language that does not require an
interpreter (ie: Mono/.NET), one that is cross-platform, and one that
stores data locally only. Forward secrecy is nice too. </opinion>

The advantage of a password manager is that you can now have one
passphrase open your encrypted password database. Individual sites and
user accounts may now use passwords of arbitrary length filled with
CSPRNG data. The closer a password is to a long string of totally random
data, the stronger it is against modern attacks.

See: https://en.wikipedia.org/wiki/List_of_password_managers

Thanks

have a nice day.yad
jdpf










Attachment: signature.asc
Description: OpenPGP digital signature

Reply via email to