1. create a local CA
2. create a certificate using that local CA
3. Then you can add the CA in your trusted authorities (Firefox does need an extra step: either enable the "security.enterprise_roots.enabled" flag, either import the CA certificate manually in it).
Details at: https://gist.github.com/cecilemuller/9492b848eb8fe46d462abeb...
"Certificate Transparency helps eliminate these flaws by providing an open framework for monitoring and auditing SSL certificates in nearly real time. Specifically, Certificate Transparency makes it possible to detect SSL certificates that have been mistakenly issued by a certificate authority or maliciously acquired from an otherwise unimpeachable certificate authority. It also makes it possible to identify certificate authorities that have gone rogue and are maliciously issuing certificates."
"Certificate Transparency logs use a special cryptographic mechanism to facilitate public auditing of certificates and logs. This special cryptographic mechanism, known as a Merkle hash tree, is a simple binary tree consisting of hashed leaves and nodes (see figure 1)."
EDIT: And you’re saying the SSH model is broken then. Also you can verify the certificate signature via another channel, like a git repo of all the signatures of most important websites (I know, it look like a CA)
1. SSH's whining about first connection fingerprint trusting is needlessly petty and nobody actually checks the fingerprints, and in many cases they have no need to do so anyways.
2. Almost all cert errors a user will encounter in the real world are the fault of misconfiguration (wrong domain) or pathological/greed-driven behavior (expiration) rather than something that actually impacts the confidentiality of the connection (which is what we care about).
3. The fact that all cert errors are treated as the same severity (red screen! exclamation points!!1 YOU ARE IN DANGER!!!1one) conditions people to click by them without thought.
> 1. SSH's whining about first connection fingerprint trusting is needlessly petty and nobody actually checks the fingerprints, and in many cases they have no need to do so anyways.
I disagree, but this is really a question of configured defaults and security UX. The first connection you make to a server is not secure, and impacts the security of all subsequent requests to that server.
> 2. Almost all cert errors a user will encounter in the real world are the fault of misconfiguration (wrong domain) or pathological/greed-driven behavior (expiration) rather than something that actually impacts the confidentiality of the connection (which is what we care about).
This is the great success of TLS - attacks are so rare that most users won't encounter them. Misconfiguration is indistinguishable from an attack, so the only reasonable thing to do is to warn the user as if it is an attack. Expiration is not a money grab, especially since the CA with the shortest expiration is also completely free. Expiration is a great thing. It limits the window of vulnerability for compromised certificates, and means that revocation lists like those shipped by chrome do not have to grow endlessly large, since expired certificates can be pruned.
> 3. The fact that all cert errors are treated as the same severity (red screen! exclamation points!!1 YOU ARE IN DANGER!!!1one) conditions people to click by them without thought.
With HSTS, that's not an option - and chrome can be configured by sites and enterprises to disallow bypassing certificate warnings. For example, try bypassing this one:
But when I spin up a new cloud server built on some image from a cloud provider, I am not sure how I can verify this certificate. How can I verify I am not being Mitm'ed on this new server?
Or maybe the cloud provider is dropping the ball in not giving me the fingerprint when I request its creation?
But yes, it does seem broken to me.
SSH actually follows the same model as SSL in this respect. It's just that basically everyone goes self-signed and there isn't a big institutional system to distribute SSH CA's.
It's more common in enterprisey environments where you have config management to distribute the CA but you can do it right now https://www.lorier.net/docs/ssh-ca.html
I mean- if there are adversaries out there trying to hack our communications- then we need to let these adversaries try so that we can engage them head on so that their methods become published, public, and thoroughly analyzed by the people in charge of strengthening our protocols.
This approach would be like fracturing the bone to make it stronger- we allow nation state hack us in order to figure out ways to stop and prevent such hacks using open and transparent software alone. Trusting any group of people anywhere "just because they're trust worthy" feels like a variable defining <the contents of its data> as equal to <the contents of its data>. It just doesn't make sense for a variable to trust itself "just because" because then I wonder if something fishy is going on under the hood.
1. Maximum cert lifetimes are falling, once upon a time you'd just pony up the cash and get five years. A year ago it was 36 months, for a few months now it's been 825 days, and there is downward pressure. So you are still going to need to renew this cert, and that means...
2. You can and should automate. Imagine buying a device in 2018 that expects you to manually input an IP address because "Eh, we could do DHCP but this was less effort (for us)". That'd be crazy right? Time to feel the same way about certificate automation.