On Thu, May 14, 2015 9:02 am, David E. Ross wrote: > With "cyberwarfare" constantly discussed in the news, U.S. Congress, and > other venues, it appears to me that government CAs should indeed be > restricted to the TLDs of their respective jurisdictions. > > Furthermore, since governments can apply pressure (often secretively) to > commercial enterprises, a similar restriction should be applied to all > commercial and non-government CAs. In this case, they should be > restricted to TLDs of those jurisdictions where they have registered and > whose governments have granted the CAs permission to operate.
Unsurprisingly, this would make online communications less secure, rather than more secure. If I operate uncomfortable-for-the-us-government.com. If the US seizes that domain (asserting jurisdiction over .com, as they have in the past), then they can also compel a US CA to issue a cert. Or just use the domain-validation case. However, if I'm concerned about my users' security, I might choose to use a Chinese CA, on the assumption that China will not blithely cooperate with the US's request. That in and of itself doesn't offer security, but I can make this more secure in several ways: 1) I can pin to that Chinese CA, *forcing* the USG to cooperate with the Chinese gov to get a certificate 2) I could pin to an EV root, forcing the USG as newly-appointed owners of that domain to go through an EV validation if they want to MITM my customers 3) I could a-priori negotiate with the Chinese CA and develop a set of authorized requestors for that certificate, with the Chinese CA's policies such that they'll review that information before issuing, even if the domain information changes There are plenty of other controls that are practical, today, to provide such damage mitigation. Indeed, they're helpful precisely when you *don't* trust your government to responsibily/securely operate the TLD space. Think about how it would be for sites like google.com.ccTLD. Any of those .ccTLDs could MITM users traffic. Under today's policies, where Google can pin users to its intermediate, that's not possible. In tomorrow's world, versions of that intermediate would have to be signed by each of the country's appropriate CAs, or be denied secure communications with that country's citizens. I think there's also the broader consideration of whether Mozilla's policy interests are served by promoting borders on the Internet, which David's proposal certainly does, but the broader question invariably does. https://www.mozilla.org/en-US/about/manifesto/ , Items 2, 4, and 6 all seem relevant to the broader discussion of the implications of such a policy. In case it's not clear, I think imposing name-constraints on CAs to be bad for the web and not a scalable solution, even if it appears attractive :) _______________________________________________ dev-security-policy mailing list dev-security-policy@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-security-policy