> The third party should be permitted access only under carefully controlled conditions, as for instance, a court order. Thus far, no system design has been shown to meet this requirement fully on a technical basis alone. All proposed systems also require correct functioning of some social linkage, as for instance the process of request for access, examination of request for 'legitimacy' (as by a court), and granting of access by technical personnel charged with access control.
The social issue is a significant problem. How do you make sure court officials are not bribed? Who has access to the key escrow service, and what sort of checks make them trusted? Employees snoop on things all the time [2, 3, 4]: how are you going to ensure this does not happen with all communications for an individual?
[1]: https://en.wikipedia.org/wiki/Key_escrow
[2]: https://www.wired.com/2008/05/five-irs-employ/
[3]: https://reason.com/2015/05/20/more-irs-employees-abuse-acces...
[4]: https://www.accountingtoday.com/news/irs-celebrity-snoop-get...
Schneier does conclude his post with a general "weakening encryption does more harm than good", but the title still uses the term backdoor. No one who's learned a bit about crypto (omitting Lindsey Graham here https://www.politico.com/news/2019/12/10/tech-companies-bipa...) is advocating for a "master key," short key lengths or backdoors that use obscurity to keep bad actors from exploiting them.
Key escrow is real & already deployed in things like FileVault's Institutional Recovery Key https://support.apple.com/en-us/HT202385
All institutions don't use the same key and the vendor doesn't have to have a copy.
[Microsoft offers companies a recovery mechanism too https://docs.microsoft.com/en-us/windows/security/informatio...]
My guess to how key escrow weakens protections is through implementation mistakes. Are there also mathematical weaknesses that effectively shorten the key length making brute force more feasible??? Why is the discourse on the anti-lawful-intercept side so imprecise?