Possible to combine smartcard PIN with key password?

[adrelanos]

Peter Lebbing:
>> By "part" I don't mean split one key in halves, but rather use two keys.
> It's an interesting thought, I'll definitely give you that. However, if you need
> that kind of protection, I don't think you should use a normal computer with a
> normal operating system. It seems to me, to attack your smartcard, they would
> need to either hack your PC, or have physical access. In both scenarios, the key
> on your hard disk is not secure anymore either.
> 
> Can you think of a scenario where the on-disk key adds security beyond the
> smartcard?

Yes.

Scenario #1
###########

Adversary capabilities:
- Can physically steal the smartcard.
- Capable of dismantling a smartcard to extract the key its holing.
[Maybe not now, but maybe in a few years the tool required to so so will
be available. Only making up the scenario here.]
- Not capable of breaking gpg's key encryption/password protection.
- Not capable of rubber-hose cryptanalysis.
- Not capable of installing a miniature camera and/or hardware keylogger.

Scenario:
- Password of gpg key has sufficient strength (length, entropy, you name
it).
- Power of computer holding the key is off long enough.
- No data left in RAM (hence no coldboot attack possible).
- Adversary hasn't done an remote/online attack, therefore the password
of the gpg key is still secret.
- Adversary obtained smartcard and hdd by using a one-time physical
attack. (For example a robber could steal a notebook and smartcard.)

Result:
- 1) smartcard only: fail [adversary extracts key from smartcard]
- 2) smartcard + gpg/encrypted/password protected key: ok [adversary
extracts key from smartcard, but fails to break gpg's encrypted private key]
- 3) gpg/encrypted/password protected key only: ok [adversary extracts
key from smartcard, but fails to break gpg's encrypted private key]

I'd prefer result 2) and 3) over 1).

###########

Scenario #2
###########

Adversary capabilities:
- Not capable of stealing/dismantling a smartcard to extract the key its
holing.
- Not capable of rubber-hose cryptanalysis.
- Not capable of breaking gpg's key encryption/password protection.

Scenario:
- Adversary compromises the victims operating system using a
remote/online attack.
- Adversary reads the password while it is in RAM.
- Adversary downloads the victim's private gpg key from hdd.

Result:
- 4) smartcard only [there is no key the adversary could download from
hdd]: as usual, key can be abused as long as smartcard PIN is cached, at
least the smartcard key can not be extracted; no further abuse possible
once detected, smartcard removed [and operating system reset to
clean/immunized state]
- 5) smartcard + gpg/encrypted/password protected key: as usual, key can
be abused as long as smartcard PIN is cached, at least the smartcard key
can not be extracted; no further abuse possible once detected, smartcard
removed [and operating system reset to clean/immunized state]
- 6) gpg/encrypted/password protected only: whole key can be extracted
and abused as long as not revoked

I'd prefer result 4) or 5) over 6).

###########

Conclusion:
We can't be sure of adversary capabilities. To archive the best possible
outcome in both, in scenario #1 and #2, it seems to be, that it's
worthwhile to combine the security properties from both, smartcards and
on key encryption.