4 ms·
Remember that you only need one of the hundreds of CAs (650 at my last count) to generate you a cert to essentially compromise a site's connection security. Fre
by peterwwillis 11y ago
Remember that you only need one of the hundreds of CAs (650 at my last count) to generate you a cert to essentially compromise a site's connection security. Free certs mean more sites will use TLS, which means there will be more targets, which means more incentive to start attacking the weakest CAs and/or their verification practices.
But this is kind of a good thing, because after enough attacks on the old model, people will ask for an improvement or replacement of the model.
- alwillis 11y agoThanks to HTTP Public Key Pinning (HPKP), this is no longer a threat: HTTP Public Key Pinning, or HPKP, is a security policy delivered via a HTTP response header much like HSTS and CSP. It allows a host to provide information to a user agent about which cryptographic identities it should accept from the host in the future. This can protect a host website from a security compromise at a Certificate Authority where rogue certificates may be issued for your hostname. You can read all about it: https://scotthelme.co.uk/hpkp-http-public-key-pinning/ https://scotthelme.co.uk/hpkp-http-public-key-pinning/ And here are web-based tools for examining and generating HPKP hashes: https://report-uri.io/home/tools https://report-uri.io/home/tools
- peterwwillis 11y agoYou don't get to say there is no threat if almost every website today is vulnerable. First, this is optional software that almost every public server in the world is currently not using. It's like saying just because executable whitelisting exists that downloaded exploit payloads on operating systems is no longer a threat. If nobody uses it, the threat still exists. In addition to actually implementing it on your server, every single user still has to make a secure initial connection over a trusted network on every device they'll ever use to get to that site. But not every client even supports HPKP. There's lots of older software which doesn't support it, and IE doesn't support it at all, which by itself would leave 12% of all clients vulnerable. https://projects.dm.id.lv/Public-Key-Pins_test https://projects.dm.id.lv/Public-Key-Pins_test https://en.wikipedia.org/wiki/Usage_share_of_web_browsers https://en.wikipedia.org/wiki/Usage_share_of_web_browsers
- alwillis 11y agoYou don't get to say there is no threat if almost every website today is vulnerable. I didn't mean that literally the threat has been eliminated due to HPKP; it's yet another tool that can be used. First, this is optional software that almost every public server in the world is currently not using. It's like saying just because executable whitelisting exists that downloaded exploit payloads on operating systems is no longer a threat. If nobody uses it, the threat still exists. HPKP isn't software--just additional HTTP headers that pretty much every web server can be configured to send. In addition to actually implementing it on your server, every single user still has to make a secure initial connection over a trusted network on every device they'll ever use to get to that site. True; it's even described in the RFC: https://tools.ietf.org/html/rfc7469 https://tools.ietf.org/html/rfc7469 Key pinning is a trust-on-first-use (TOFU) mechanism. The first time a UA connects to a host, it lacks the information necessary to perform Pin Validation; UAs can only apply their normal cryptographic identity validation. (In this document, it is assumed that UAs apply X.509 certificate chain validation in accord with [RFC5280].)
- Gregordinary 11y agoIt isn't foolproof, user installed Root CAs or roots installed by 3rd party applications are given permission to override pins. So HPKP wouldn't protect against scenarios like Superfish. https://www.imperialviolet.org/2011/05/04/pinning.html https://www.imperialviolet.org/2011/05/04/pinning.html