[talk] passwd entropy and strength

Jim B. jpb at jimby.name
Mon Nov 6 15:10:21 EST 2017


* Thomas Levine <_ at thomaslevine.com> [2017-11-06 11:22]:
> I am very satisfied with my password management approach, and
> I keep debating whether it is safe to share. On one hand,
> I theoretically think that sharing it would not provide an adversary
> with enough information to guess my passwords or otherwise bother me,
> but I also wonder whether it would tell the adversary enough to study
> the relevant approaches/software enough to come up with a vulnerability.
> Does anyone have any commentary on my concern?
> 

Like many valuable things, it depends on your risk model.
Who is your adversary?  A 3LA?  Large well funded criminal
enterprise?  A really good hacker?  Random scripty?

For most folks, I suspect it is the latter.  On the other hand,
it doesn't matter who it is - personally I want to keep my shit
safe from *all* prying eyes.  It may not be great shit but it's
*my* shit, and I want to keep it that way.

You could of course, follow the web of trust model - share
individually with those you trust and gain feedback to improve
your defenses.

That said, I'll share my approach.  Just don't hack me and steal
all my shit :-)  I'm going to number all the elements below
to allow for easy commentary.  I welcome all comments and
suggestions on how to improve this setup.


1. My "password vault" is just a text file on my TrueOS laptop.

2. The file (in some directory)  is encrypted with gpg2(1) with a
   strong password.  I do not keep a clear text version on hard
   disk.

3. To edit the file, I create a memory disk, format it with newfs(8)
   mount it, and lock down the mountpoint.
   
4. I copy the encrypted file to the memory disk, decrypt, edit,
   save, and re-encrypt it all on the memory disk.  Lastly, I
   use "rm -P" to "securely delete" the decrypted copy that
   I edited.

5. The encrypted file is copied back to my hard disk.

6. The memory disk is unmounted and deleted.

7. The ASCII-armored file is then steganographically embedded
   inside a JPEG file.  I use outguess(1) (by Niels Provos)
   for the steganography operations.  This operation has a
   completely different password.

8. The .JPG file is copied to a secure location online.


Thus, I have a local ASCII armored encrypted file on my disk,
and I have a copy of that file embedded inside a JPEG file
on my disk, and also somewhere online.


9. If I just want to view the file, I open up a new terminal,
   cat the encrypted file and pipe it to "gpg2 -d" which
   outputs the clear text to the terminal window.  I
   copy/paste whatever I need, and delete the terminal window
   when done.  pinentry(1) is used for passing the password
   to gpg2 in this step.


That's it.

Comments welcome!
Jim B.

 




More information about the talk mailing list