Let’s Encrypt has been operational for about two years now, although the project originally began in 2015. Let’s Encrypt is the saving grace of HTTPS, but exactly because it is the saving grace of HTTPS is the reason that I dislike its endorsement.
Suppose that tomorrow, a security researcher discovers a critical flaw in CertBot or some other part of the Let’s Encrypt certificate issuance system, and in a week, almost every Let’s Encrypt cert is going to get tossed into the CRL, with no ability to create new certs.
They couldn’t do it. They couldn’t possibly toss 100 million certificates into the fire, because LE has already reached a point where it is too big to fail. You can’t tell your users, who expect their website encryption to come for free, “Hey, your CA got compromised, so you’re going to have to pay $20 or more for a cert from Verisign, GeoTrust, or Comodo, because there are no other free, secure CAs available. Sorry.”
And if it comes to that, two things happen:
- Verisign et al. gouge prices and have the biggest cert bonanza ever, because website owners have no other choices.
- An HTTPS blackout happens, and half of all HTTPS-enabled websites have no choice but to fall back to regular HTTP. And if this happened with a version of Chrome where insecure browsing is banned, then you can just forget about that website unless you are a website owner and choose (1).
You have to remember the situation before Let’s Encrypt: Browser vendors, most especially Google and Mozilla, were pushing as hard as they could toward eradicating HTTP and enforcing HTTPS everywhere, in light of the Edward Snowden and NSA hysteria-bordering-paranoia. However, SSL/TLS certificate options were limited at the time: existing free certificate services had been founded long before then and were commonly suggested for people who were absolutely desperate for a free certificate, but were nonetheless unpopular among CA maintainers due to rampant abuse. In other words, on the idealistic side, people believed that every site ought to have HTTPS. But on the practical side, they asked if your site really needed HTTPS if you can’t afford a certificate and you are just serving static content.
Today, those old free CAs have been abandoned by CA maintainers in favor for the one CA to rule them all: the ISRG/Let’s Encrypt CA. I mean, we’re obviously not putting all our eggs in one basket here – if something goes wrong, we still have hundreds of CAs to go by, and if an owner really needs their HTTPS, they can just shell out $100 for a cert. That’s right, if you’re a website owner who cares more about their website than the average Stack Overflow user, you should really consider shelling out money, even though we’re sponsoring a cert service that is absolutely free! Oh, and if something goes wrong, you get what you paid for, right? My logic is totally sound!
Let me reiterate: in the case of a future catastrophe, assuming that we are enough time into the future that browsers have placed so much trust in the HTTPS infrastructure that they now put prevent casual connections to insecure HTTP websites, there are two answers based on how much money you have:
- You’re f**ed, along with millions of website owners. More news at 11. Maybe the folks at Ars Technica can tell you what to do. Except they’re also too busy panicking about their personal websites.
- Buy a cert before they raise their pri– oh, too late, they’re $50 a pop now.
So, I think the problem at hand here is the philosophy behind trust. Trust is such a complicated mechanic in human nature that it cannot be easily automated by a computer. When we make a deal on Craigslist, how do we know we’re not going to end up getting kidnapped by the guy we’re supposed to be meeting with? Is the only reason a bureaucracy trusts me as an individual because I can give them an identification card provided by the government? But how can I, as an individual, trust the bureaucracy or the government? Only because other people trust them, or people trust them with their money?
How does this tie into the Internet? How can I trust PKI, the trust system itself? What happens if I tie a transactional system – specifically the likes of Ethereum – into a web-of-trust system such as PGP? What happens if I tell people, “vote who you trust with your wallets“? What is a trustable identity in a computer network? What remedies does an entity have if their identity is stolen?