Slashdot Mirror
Let's Encrypt Is Now Officially Trusted by All Major Root Certificates (bleepingcomputer.com)
Let's Encrypt has announced that it is now directly trusted by all major root certificates including those from Microsoft, Google, Apple, Mozilla, Oracle, and Blackberry. With this announcement, Let's Encrypt is now directly trusted by all major browsers and operating systems. From a report: While Let's Encrypt has already been trusted by almost all browsers, it was done so through intermediate certificate that were cross-signed by IdenTrust. As IdenTrust was directly trusted by all major browser vendors and operating systems, it also allowed Let's Encrypt to be trusted as well. With Let's Encrypt now being directly trusted, if there is ever a problem with IdenTrust and they themselves become untrusted, Let's Encrypt users will still be able to function properly.
Gee, now if the certs would last longer than Trump's attention span, Let's Encrypt could actually become useful. At this point, they should rename it "Let's Momentarily Encrypt."
Trusted by root certificates? That is not how root certificates work. Bad article and bad headline for a tech site
hi! cool.
Microsoft? Check.
Google? Check.
Apple? Check.
Mozilla? Check.
Oracle? Check.
Blackberry? Che... wait, what?
Netcraft confirms it, this list is dead.
Let's Encrypt has become a single point of failure for the majority of web sites and is thus too big to fail, which is a terrible thing to say about a CA. It needs to be broken up or preferably a more distributed way of securing web site encryption needs to be established.
Let's Encrypt is a really good setup for people who want to learn how to automate their system. While free and easy to set up (it took me about an hour to get https on my websites with it), the certificates only last 90 days, with the justification being that people should learn how to automate things.
Since I have multiple redundant nodes which I rsync to, I had to use the --manual-auth-hook option to certbot-auto to push the challenge-response tokens Let's Encrypt uses to authenticate website. I also use Ansible to log in to all of my nodes to update the certificates once they are generated.
Note that Let's Encrypt does log the IP of the machine used to generate the certificates; while these IPs have not been made public, the EFF keeps threatening to do so, which causes some lively discussion on the Let's Encrypt forum.
*ALL CA* are a single point of failure, it is not just let's encrypt
Higuita
... why the vitriol on bleepingcomputer and why msmash likes it so much. As does BeauHD.
Let's Encrypt has become a single point of failure for the majority of web sites
I generally think of "single point of failure" as one thing breaks and it immediately takes everything else down with it. With certificates, you should be renewing them 30 days before they expire. If Let's Encrypt suddenly ceased to exist, you would have 30 days notice that they are gone, and thus 30 days to switch to a different certificate provider and continue on with zero downtime. That's not my definition of single-point-of-failure. So it's really only a single point of failure for websites whose admins can't be bothered to monitor their processes, and can't be bothered to read tech-related websites and blogs (as something like that would be posted about everywhere).
And we have known how to fix it since about 1988-1990 (PGP), before HTTPS was even a thing. Our entire CA system was obsolete before we started using it. Hopefully, some day we'll upgrade to 1990 tech and then identities will have multiple parties certifying them.
As copyright owner of this comment, I authorize everyone to defeat any technological measure which limits access to it.
This is a boneheaded rationale. It means that the risk of the process (automated or not) breaking arises more often than it needs to.
Many users want encryption that is in place and remains in place, not encryption that stops working, or potentially stops working, at very short intervals. Let's Encrypt produces a series of short-interval risks of failure. Not going to bite on that.
Came here to say the same thing. The headline makes no sense whatsoever.
From the official announcement: "While Let’s Encrypt is now directly trusted by almost all newer versions of operating systems, browsers, and devices, there are still many older versions in the world that do not directly trust Let’s Encrypt. Some of those older systems will eventually be updated to trust Let’s Encrypt directly. Some will not, and we’ll need to wait for the vast majority of those to cycle out of the Web ecosystem. We expect this will take at least five more years, so we plan to use a cross signature until then." So let's not hurry with the celebrations. It will take 5 year at least to happen ......
Let's Encrypt has become a single point of failure
How so? You do realise there are systems in place to handle faults in certificate issuing processes, and outside of the issuing process they are not in any way involved right?
Before you declare something a single point of failure and a major drama, maybe define what the failure mechanism and the consequence is first.
Letâ(TM)s Encrypt is a free, automated, and open certificate authority brought to you by the non-profit Internet Security Research Group (ISRG).
So if you need an SSL certificate for cheap, you can go to them. https://letsencrypt.org/
Don't fight for your country, if your country does not fight for you.
You say you're arguing against it being a single point of failure, but your actual argument is that one of the many ways it can fail and completely break, happens to be easy to avoid.
If we hypothesize that expired signatures were the only way that PK cryptosystems fail, this might actually be a good argument. I propose that everyone who is unhappy with the overwhelming weaknesses and variety of vulnerabilities in todays PKI should all move to your universe, away from this sketchy shithole where CAs and governments have already been caught many times, fucking up or being bad actors. That way all the problems people have been thinking about for the last 3 or 4 decades, just magically go away! What's not to like about that? Anyone see a problem with my proposal?
the PRISM list.
Domestic spying is now "Benign Information Gathering"
These guys did something right and I applaud them. Much better than managing your own certificates and getting your users to accept them.
You mentioned the Wikipedia article "Web of trust". It acknowledges that getting your key signed for the first time is impractical for many. True, a key signing party will help your key become trusted in the same village. But that doesn't help you build a robust set of paths through the web of trust to users on the other side of the planet unless several people who attended the same key signing party also routinely travel internationally to key signing parties in other countries. And with the U.S. TSA and other national air travel regulators ramping up their security theater in response to terrorist threats, international travel has become more impractical over time. The terrorists have won.
*BSD uses Mozilla's root certificate bundle.
FreeBSD The port port security/ca_root_nss provides Mozilla roots. (Source: chatwizrd's post). NetBSD The package security/mozilla-rootcerts provides Mozilla roots. OpenBSD This libressl commit states that OpenBSD's LibreSSL library provides Mozilla roots.So establishing the web could be somebody's job. Imagine if i walk into a AAA storefront, show them my ID and pay a small fee, and they sign my cert.
My bank could do the same. or 711 for that matter. Hell, the DMV ought to, establishing identification is half of their job anyway.
All i'm saying is, we could have more "web-of-trust" infrastructure then just key signing parties.
English speakers use "one" or "someone" as the impersonal pronoun. For example, your comment could be reworded as follows: "Someone who can't set up a cron job shouldn't be in charge of an Internet-facing machine altogether."
So establishing the web could be somebody's job.
I believe that job is called a notary. And you're right that a notary firm operating in multiple villages would have the resources to build the web beyond one geographic area.
Hell, the DMV ought to, establishing identification is half of their job anyway.
Even so, good luck getting that, or any other new duties of the DMV, past the minarchists in the Republican Party of each U.S. state.
There have been situations where a CA was deemed untrustworthy. Some have been removed from browsers and operating systems. Of course, when that happened there were "systems" in place to ease the transition for clients of these CAs. But Let's Encrypt is fundamentally different: Let's Encrypt certificates are free and this has led to many web sites getting certificates that otherwise would not have enabled TLS or used fewer domain names for cost reasons, etc. Some of these sites (a lot actually) have locked their sites to HTTPS through long-lasting HSTS declarations. Even if the site owners themselves haven't tied their web sites to HTTPS, the expectation created by the availability of Let's Encrypt and the subsequent changes in browser behavior have. And you can't just tell the owners of these sites to switch to a different free CA: it doesn't exist. If Let's Encrypt somehow breaks or goes away, it will be a serious problem for lots of web site owners. And there's also the problem that migrating more than half of all web sites away from a CA takes a long time. This makes Let's Encrypt "too big to fail".
And yet people are able to get X.509 certs signed, and we even have things like LetsEncrypt. The evidence suggests getting signatures isn't really all that hard, since 100.0% of the websites that implement HTTPS somehow managed to do it.
So why stop at 1? The only people who come out ahead by us having single point of failure, are the attackers. I think we should move from a pro-attack to a pro-defense strategy, though I guess we should let the people at NSA, FSB, Chinese government and the Mafia weigh in on this before we make any hasty decisions.
I linked to the WoT article to inspire/remind people to think about the robustness of multiple parties attesting to an identity instead of just one, as well as how you decide how much to trust any one given CA. (Which is something nobody does today.)
What if one of the many signatures expires?
What if one of the many certifiers disagrees with the others, due to malice or mistake?
The WoT beats the living shit out of what we're doing today. It degrades gradually and more slowly when faced with simple failures, and it requires conspiracies (instead of someone coercing one single party) to undermine it. Perhaps that's why we don't use it: because it would be more secure, inconveniently too secure when you need to spy on someone. Or perhaps it's because people want to pretend that your confidence is either 0% or 100%, in spite of the fact that nothing ever really works like that.
Actually some people do that (an international path through the WoT isn't that uncommon) but you're right that what happens today in PGP's WoT often isn't enough, and it really wouldn't be enough for everyone.
But I wasn't suggesting that the faceless companies that you currently fully trust (hey someone, remind me: why?), have to be left out and replaced by amateurs, as somehow turned out to be the case with PGP. If we implemented the web's PK like PGP did it, then you could still have your cert signed by Verisign and LetsEncrypt and Comodo and your neighbor and your bank and state government and those people you lifted pints with at the conference bar. Sure beats having a single point of failure. Imagine your LetsEncrypt signature expired then. Imagine Comodo fucked up again. Imagine your own government told Verisign to lie or your neighbor was trying to MitM you. Instead of these being disasters where thousands of people have to scramble to minimize downtime, it would be a minor nuisance, detected quickly ("hey, one of these CAs disagrees with all the others..."), and with reputation ramifications.
The catch is that we'd have to start valuing defense more than attack.
As copyright owner of this comment, I authorize everyone to defeat any technological measure which limits access to it.
DMV? Good one. I just had the experience of 'proving' something to the DMV (NY). I needed to provide 2 'proofs of residence'. My mailing address is a PO box, as the wonderful USPS does not deliver to homes in our town. One of the proofs I had was my water/sewer bill. The bill has 'YOURTOWN WATER/SEWER DISTRICT' printed across the top, and had my address (street and house) listed as 'service to property'. The genius at the DMV would not accept that, because the 'service address' did not have the town listed. Exactly what town do they think 'Yourtown water/sewer district' serves?
But it gets better. They gave me a form, which could be used to 'prove' my address. This form could be filled out by anyone, including my spouse, saying that I lived at the address I said I did. The person filling out the form doesn't have to appear in person, and the form doesn't even have to be notarized. Not sure how that proves anything.
Oh, and another form of 'proof' that they will accept? 'A computer printed pay statement'. Man, who could ever forge one of those?
Let's Encrypt has become a single point of failure for the majority of web sites and is thus too big to fail, which is a terrible thing to say about a CA. It needs to be broken up or preferably a more distributed way of securing web site encryption needs to be established.
Um, the server software is open source:
* https://github.com/letsencrypt/
Have at it.
I'm a sysadmin at a government website --- we were explicitly told by the local network police that we can't use Let's Encrypt (which was working flawlessly for over a year) because the certs expire too frequently.
Yeah, I had to run one command every 6 months ...
Now we use godaddy, if you can believe it.
Gov't agency name omitted to protect the clueless
For 'Let's Encrypt' to open the torpedo tubes.
Whoever complains about https has never lived in a country where every ISP injects adverts into every non-https website.
There are huge positive effects for the minimal trade off of having to setup a certificate. Also the more the google pushes https-only, the easier the buildmywebsiteonline.coms will make the process for the normies.