2 ms·
The thing that gets me in all these discussions is people acting like we can't build a protection mechanism that stops A from getting in but not B. Yet, these s
by nickpsecurity 11y ago
The thing that gets me in all these discussions is people acting like we can't build a protection mechanism that stops A from getting in but not B. Yet, these same people in another context say we can using words such as VPN, port-knocking, TLS, SSH, and so on. We do it all the time with some solutions being quite strong. If anything, a robustly-designed R.A.T. or escrow is nowhere near as threatening than the average app or network protocol on endpoint: practically backdoor generators in practice and their 0-days are the main way High Strength Attackers hit targets.
So, first we should stop with the double standard. There exist mechanisms that work well enough in practice which allow selective access to systems (aka front-door). There's also decent work in limiting such access, auditing it, escrowing, and so on. NSA also already does the escrow for its Type 1 encryption products with all key material managed and protected by their EKMS system [1]. If it got visibly breached, we'd probably have seen evidence of it. So, I'll skip straight to the other issues that popped up in peer review when I published a few attempts at a secure solution to the L.I. problem.
Will it have security problems in general? Yes. Damage can be limited in a number of ways, esp if scheme is decentralized. Yet, it adds risk on top of what we already given even rigorously-built solutions have had issues.
Will there be authentication forgeries for LEO requests? Yes. This might be a one-off, quickly detected thing. It might be unlimited depending on how the breach goes. Sender and receiver must be protected across the board. LEO's track record isn't good. High risk here.
Does the government have a history of abusing its access and deceiving about its usage? Yes. A great reason not to do L.I.
Will the government agree to a L.I. scheme with a strong foundation for security that ensures read-only access for a limited period and can't subvert system? Clive Robinson suggests that even a theoretically, bulletproof L.I. system won't be acceptable if it limits them and it will be forced to be more "flexible" by their sway over courts/lawmakers. Seems true given current situation.
Can they prevent attacks via malicious sys-admins or infiltrators? Snowden says no and with quite the credibility. ;) High risk here.
And the big one: will the KEYMAT-holder survive the inevitable attacks of all intelligence services, hackers, and organized crime combined? And all the things they will do to the organization or its employees? Gravely, high risk here.
Dirk Praet brought up the last question in a discussion where I had tried to cover every single angle in design, personnel, and ops. In the end, Dirk and Clive are right: the government will force the system to have too much privilege, centralization, and/or accessibility; data will be mass compromised in a free-for-all by the opponents. Whatever concrete requirements come out of such legislation will present vastly higher risk than with key escrow or remote administration tools in general.
So, there's extremely high risk in warrant process, key storage even with HSM's, and organizations administering this even at SAP security level. More risk than any DOD, NSA, or FBI have faced before given the concentration of enemy effort that will occur. Their track record on smaller things shows they can't handle something that big. Even if they were honest as Boy Scouts, the risk assessment argues against the L.I. option 100% and regardless of what level of INFOSEC our mechanisms ensure.
Also, if they were Boy Scouts, they couldn't think like the attackers enough to stop them. Proposal is untrustworthy whether they are or not. That's No^2. ;)
[1] https://en.wikipedia.org/wiki/Electronic_Key_Management_System https://en.wikipedia.org/wiki/Electronic_Key_Management_Syst...