Re: What is the point of a warrant?
@Paul Crawford:
Not quite sure what you have in mind and as I mentioned it can get complicated. With the proviso that the third party can be trusted due to the fact that it can be extended to as many different parties as required to be secure and that the PKI need not be limited to a single type and although significantly more involved it is possible to accomplish the same thing with conventional keys or even one time pads:
I have a message I wish to remain secret. I prepare it on a secure system and send a secure message to my trusted third party requesting a one time public key whose private key is known only to the third party. I encrypt my message on the requested public key and then encrypt an envelope containing the third party supplied public key with the receiver's public key. The original message is now gone and there is no way to recover the message without both the recipient's private key and the private key known only to the third party.
Details can get pretty hairy, but suffice it to say that it is possible, if needed, to make it so the third party actually cannot divulge the necessary key without the active permission of the sender and the receiver and an arbitrary number of nth parties if needed.
Security can be a PIA. If you want to secure something on a password and have reasonable confidence that it remains secure as long as the password is not known, you need to come up with a long password whose characters are effectively random. Something like this that has not been published (ie not this actual one because it is compromised now): MKMKtrsquRXKogec_zuxgKRfJmHQIoQW. That should give a nominal 192 bits and likely about a good 90 bits of real security against attack; simply not guessable in any reasonable amount of time. Unfortunately, it is so awkward to use such a thing in practice that it would not likely be used.
The reason for the above is to make it apparent that there are different levels of security available at the expense of given levels of inconvenience. Security is possible in a password, but inconvenient. You would not normally do that, but you might if the need was great enough. Similarly, to ensure that a scheme like the above was more secure against attack you could make it so that access to a particularly sensitive message was only available for a limited time beyond which it disappeared entirely. That way, particularly sensitive communications could vanish forever before anyone had a chance to beat the passwords out of you. This would be pretty inconvenient, but a lot more secure.
I have little doubt that schemes capable of securing systems can be built as long as we can build systems secure against things like side-channel attacks and we can trust the hardware. I have even less doubt that current systems do not approach anything like a level of security that even a duffer less skilled than me could put in place. Any of the big players like MS, IBM, Google, Facebook, Apple, HP, Oracle, etc, etc cannot possibly be trying in any meaningful way to secure their systems. This stuff can get pretty complicated pretty fast, there are gotchas everywhere and even experts who I trust have tried will make mistakes. However, virtually every barrier to entry on to our networks has been lowered to the point that even attackers with modest resources can mount a successful attack.
I have to do a search to see if such a thing has been patented already, but while writing this up I thought of a hugely amusing invention to cure shoulder surfing and related surveillance that had been a real puzzler for me.