Implementation issues, security holes
If you want security, you have to work at it.
If you have PGP (for instance) installed and use it, your E-mail is
probably safe after you send it, and incoming mail is safe until you
get it (apart from the dangers of mail being lost, the pattern of mail
analyzed and so on).
But consider the following attacks on your mail's privacy:
- An attacker wanders into your office during lunch and sees your
letter on the screen while stealing your wallet
- An attacker discovers that you are using an X display with
"xhost +" (open access), starts a grabber that records all your
keypresses, and uses that to unlock your secret keyring
- An attacker attaches to your local network, discovers that
you are using PGP on a diskless PC, and that the plaintext is
swapped to the fileserver while you are decrypting it
- An attacker with access to any directory in your PATH installs
a rogue PGP version that always adds the session key encrypted
for an extra recipient: the attacker. (Do you have the current
directory in your Path?)
- An attacker mails you a Word document that installs a macro in
your templates which will send an unencrypted copy of the document
to an attacker every time you invoke the "file|mail" menu in
Microsoft Word
- An attacker aims a high-powered light detector through your
window at the wall behind your back, and is able to reproduce
the picture on your screen from the timing of the CRT emission
- An attacker discovers that exmh (an Unix mail agent) leaves the
decrypted files in /tmp while you are reading them, and uses
superuser privilleges to read them.
- An attacker roots through the dustbin outside the building and
discovers the printout of the message that you forgot to put
into the document shredder
All these are relatively simple attacks that do not depend on cracking
any part of the actual cryptoalgorithms involved. If you are serious
about secure E-mail, you should consider these issues too.
Harald.T.Alvestrand@uninett.no
Last modified: Fri Nov 3 10:40:35 1995