Least worst golden key
tedunangst.com
tedunangst.com
I disagree with this premise. I think it's dangerous to publicly and unashamedly discuss how to build a broken system like this, because it reduces the cost of performing that design in the future.
For example, if you run an online service, you should spend your time building a system that makes it impossible to spy on your users, not writing an internal procedure that describes how you would spy on your users if someone forced you to.
That said, I think the best argument against escrow for those who trust the FBI is that it will chill innovation and free speech, and push consumers and businesses to purchase from companies that don't do escrow (because they're outside the jurisdiction of whoever does it).
Imagine if someone asked for a way remotely administer a nuclear power plant over the Internet. It's possible to build a reasonably secure system -- VPN, one-time passwords, company-issued hardware, etc. But security isn't exactly intuitive to a lot of people and there are a bajillion ways to screw this up. So given the dangers involved with a nuclear power plant, the smarter approach is to simply insist that nuclear power plant control systems be air gapped from the rest of network rather than to figure out a 'good enough' way to secure remote access.
How about a system where you need n of m pieces of data to obtain a key, for suitably large n < m. Then you can split your key into m subkeys, and give each one to a different trusted third party, who would only reveal the key if presented with a warrant.
Then in order to decrypt the data, the government would have to subpoena, say, 20 out of 40 separate public organizations. It wouldn't be an easy thing to do, so hopefully they wouldn't do it often (and we would know if they did), but in extreme cases where someone's life depends on it, and everyone in the world agrees that it's necessary, it could still be done.
This system could also be made difficult to compromise if every week or so you would generate a new set of subkeys and send them out to the secret keepers, who would then delete the old keys they had for you. Then even if someone got one of the key stores, it would become useless after new keys were sent out, since all the other keys necessary to use it would be permanently deleted. To steal a key, someone would have to compromise all n of the key stores at the same time.
Something like this may even be useful for things that you want to be able to recover yourself. Say I have my bitcoin wallet stored securely on a flash drive, but if I lose it, I can get it back through this method by verifying my identity to say five organizations that I trust. Does anything like this exist already maybe?
It's really neat, actually and can be efficiently and easily implemented just going from wikipedia's description.
On which side are you on? Seriously - someone (WaPo) brings up a totally unaccecptable method and then someone brings up a slightly less unacceptable method. This sounds to me exactly the way some politics scheme works.
"There’s my least worst idea. A per device backup key, split in pieces, stored on paper in underground silos."
Again, no thanks. Any form of key escrow is a backdoor.
Be careful with your wishful thinking.
You want a least worst escrow? How about an escrow where the keys aren't written down. How about you just restrict your salt to N, N+1, .., M and have the TLAs brute force it? Actually trust no one.
Just choose a key strength that costs about $10,000,000 to brute force it, and keep increasing key strength as more efficient methods are found.
Wasn't that the approach in the early days of computer crypto? Where keys bigger than X were verboten?
That's what I want to see happen. Key escrow/giving law enforcement a foot in the door of our personal lives? No thanks.
"But what about criminals? And TERRORISTS?!"
Law enforcement has already been pretty effective at solving crimes without these capabilities. Statistically, terrorism is a minor risk.
(I would also like this to apply to end-to-end authenticated encryption between parties as well.)
the simple truth is that any back door, whether secured with a golden key or not can be broken in to
"Look, guys, I'm not saying New Earth Creationism is right; I'm just saying we should have all our scientists talk about it and find arguments in favor of New Earth Creationists."
It's not well written.
It doesn't propose a good system.
And yet sadly it succeeds in advancing the bad side of the argument, just by spreading FUD.
I'm not saying it wouldn't be a good use of money, but it might make it difficult to build support for the idea. A cheaper solution which preserves much of the security would be to use cryptographic key-sharing rather than physical sharding. The escrow keys would still be created uniquely on a per-device basis, but now Apple, the FBI, and the Justice Department must each provide a fragment of the key. Perhaps the system could be designed to release keys only when provided with a valid warrant requesting a specific key... but this would probably require warrants to be better digitized.
I understood that was the whole point.
If you're going to build a key escrow mechanism that isn't prone to universal, zero-effort abuse, there are going to be all kinds external factors that need to be considered.
Ted's article lays out the requirements to show just how bad the idea of key escrow is, including the measures that would be necessary to make it acceptably insecure. (Key escrow is - and remains - defective by design.)
The problem is compounded by the fact a golden key scenario only works with the current web services model we've become dependent upon. What happens when encryption is the default everywhere all the time and computation is performed by blind services or smart contracts mapped together over a trustless network a la bitcoin style? Everything about the experience could feel the same with the exception being control over your data is shifted from a third party to you.
One possible solution would be to establish computational governance. At least at that point you have a communal opt-in which gets you away from the problem of perspective. In the same way all motorists agree to a set of rules for a given area, you could be given access to use services if you agree to a set of policies. One such policy would be to have access to your data shared to a third party if a reasonable process is followed (a dumb example would be if a majority of your peers agree that you should be investigated). Probably another can of worms but at least then we're operating from a workable premise.
No, let's not.
Key escrow is the magical happy pill for people who think that governments and other "trusted" organizations are incapable of:
* creating bad laws
* harboring rogue officials
* making mistakes
* becoming tyrannical, dictatorial, abusive, or corrupt