[supriyo-biswas]

For anyone who’s about to say that surveillance isn’t the point of this legislation: it definitely is; we very recently saw Germany trying to MITM jabber.ru users[1], having a CA that can be asked to issue any certificate is definitely something that’d be used for surveillance purposes.

[1] https://notes.valdikss.org.ru/jabber.ru-mitm/

[Jensson]

eIDAS exists since there are many conflicting standards for electronic certificates. eIDAS is an effort to unify those standards. Maybe the clause where they say browsers has to add specific CA's is for spying, but eIDAS in general isn't to help spying its just there to help unify all the different electronic certificate services in EU.

For example banking, signing official documents like grades from school etc, all of those usecases are a part of eIDAS. That is the core of the standard and there you really want to see all the certificate information to be sure it is the right origin, since unlike browsers there is no list of trusted CAs, you just see that some organization accepted it.

Edit: Browsers already had their own standard that they think is better than eIDAS, so they don't want this to apply to them. But Occam's razor says that EU just added "and browsers should also do this" instead of there being some conspiracy behind it, it was simple to just add everything instead of leaving just browsers out.

[ngrilly]

> eIDAS exists since there are many conflicting standards for electronic certificates. eIDAS is an effort to unify those standards.

Did we need laws to "unify" all the standards we successfully use today, like IP, UDP, TCP, HTTP, TLS, Certificate Transparency, HTML, ECMAScript, CSS, DNS, DMARC, DKIM, SSH, etc.? Laws are not the right tool for this. And law makers don't have the necessary expertise.

[sshine]

It’s either laws or market forces, both have drawbacks.

While eIDAS seems like a great idea to coerce member states into adopting a common standard, it just also happens to sneak EU-centralist ideology in, and total digital surveillance is the 0th application of that ideology.

The big catch with EU is: once you opt in, opting out is very difficult.

[troupo]

There are also great many standards we use today that were unified and enforced through laws.

Open any law on produce, construction, cars, industrial equipment (and a million others), and you'll find thousands of specs and standards mandated by law, and for a reason.

[Serenacula]

I think ECMAScript my actually be a counter example, no? Isn't that also governed and funded by the European council?

[smarnach]

There definitely isn't a law mandating Javascript engines to follow the Ecmascript standard, which would be the equivalent of what's happening here.

[supriyo-biswas]

> Browsers already had their own standard that they think is better than eIDAS

Unlike the Browser/CA forum rules which are security focused, EIDAS comes from a government mandate first and foremost, so the concern isn’t entirely subjective as you suggest.

> "and browsers should also do this" instead of there being some conspiracy behind it

The law isn’t RFC 2119 where there is a distinction between SHOULD and MUST: the law is all about what an entity MUST do, so bringing up “should” in this context isn’t helping the point you’re typing to make.

[Jensson]

I don't get what your point is here, you said this and that is what I argued against, your points here does nothing to defend this: "For anyone who’s about to say that surveillance isn’t the point of this legislation".

> Unlike the Browser/CA forum rules which are security focused, EIDAS comes from a government mandate first and foremost, so the concern isn’t entirely subjective as you suggest.

I didn't say this was subjective. My argument was that it is easy to see why EU would do this without having surveillance in mind. They just wanted all certificates to follow the same standard, the main part of these standards were document signing and they thought web sites are documents so we add them as well to the standard.

> so bringing up “should” in this context isn’t helping the point you’re typing to make.

I didn't make a distinction between should and must there, that wasn't my point at all. What was hard to understand there? This bill is first and foremost about document signing, and then they added a clause that it also applies to browsers. That is the main part of my argument.

A bill that first and foremost targets document signing doesn't seem like it was obviously made to add spying on browsers, if that is what they wanted they would have labeled it "web protection bill" or something like they did with the chat one, they aren't afraid of saying it is about spying when that is what they want.

[miohtama]

More healthier approach for the EU to get e.g. the document signing to a single standard would be

- Make sure there is an open standard (is there?)

- Fund and promote its open source development

- Have an industry lobbyist non-profit to onboard individual businesses

If the goal is to ”promote standards” the way this is being done does not seem to be aligned the 50 years of software industry standard development, with the examples like TCP/IP, PNG, AV1 and so on.

[logifail]

> signing official documents like grades from school

I have no Earthly idea why a) this needs to be done digitally, or b) for the EU to be involved (at EU level) with this.

Unfortunately if you pitch mission creep vs the principle of subsidiarity, the former wins every time.

[willeh]

University grades are standardised already. This is useful because it allows people to work in other countries, digitally signing them prevents fraud.

This is just one use case for eIDAS, then you have things like interacting with different government institutions, banks, et cetera, et cetera.

There are a lot of people who live in/work/visit other EU countries as is their near absolute right. We should therefore standardise technology on the EU level to make their lives easier.

[logifail]

> University grades are standardised already

... for some value of "standardised"?

UK[0]: First, 2:1, 2:2, Third

Germany[1]: 1 to 5

France[2]: "on a scale from 0-20"

<chuckle>

[0] https://www.imperial.ac.uk/students/success-guide/ug/assessm... [1] https://www.uni-passau.de/en/international/coming-to-passau/... [2] https://u-paris.fr/en/higher-education-in-france/

[willeh]

Since you obviously ignorant of how it works. When you get a degree you get a transcript where all local grades are translated to to ECTS, which you then would use to apply for jobs. Of course in the tech industry grades or even whole degrees are generally disregarded but in finance and other fields they of course, are.

https://en.wikipedia.org/wiki/ECTS_grading_scale

[logifail]

> When you get a degree you get a transcript where all local grades are translated to to ECTS, which you then would use to apply for jobs

https://www.google.com/search?q=%22job+application%22+%22ECT... gets me only a handful of results and a warning that 'It looks like there aren't many great matches for your search'

Do (m)any European employers know about this scheme?

[nulld3v]

Great, very good! Now if you want to standardize encrypted communication, please do it with the help of security researchers, not like this.

[willeh]

Other than this questionable browser CA thing, do you think there are any specific flaws with the crypto system presented in eIDAS.

[nulld3v]

Alright, so I am not a security researcher so actual security researchers may not share my views. Also, as mentioned in the site, the full text of the new regulation is not public yet. And finally, I have only skimmed whatever text is available given that it's over 100 pages and I skipped over most of the EDIW stuff (it's a really complex system that I can't understand/audit in 20 mins).

But with that out of the way, no I don't have any other complaints, I think the regulation is generally a move in the right direction.

[hulitu]

> since unlike browsers there is no list of trusted CAs,

This Trusted CA is such a lie. I mean we all know that Google, MS etc does ugly things with user data but apparently we have no objection to trust them with cryptography.

[fuoqi]

A proper solution for MitM is mandatory independent certificate transparency, not outright denial of national CAs support in browsers. A German National CA should not be able to issue certificates for .ru in the first place and having a clear record of misbehavior in CT is probably not something operators of such CA would like to have even when pressured by intelligence agencies.

Browsers should get their shit together and add proper support of domain-limited CAs and add optional whitelisting of CAs for given websites.

[agwa]

> Browsers should get their shit together and add proper support of domain-limited CAs

They do in fact support this - e.g. Mozilla trusts KamuSM only for .tr [1], Chrome limited ANSSI to French TLDs [2].

However, there is no indication that the EU would be willing to accept such constraints on their national CAs. If you look at several of the current national European CAs, they routinely issue for generic TLDs like .com.

[1] https://groups.google.com/a/mozilla.org/g/dev-security-polic...

[2] https://security.googleblog.com/2013/12/further-improving-di...

[lambdaone]

Cool. Domain-limited CAs are a really good idea, and they don't need anything like dynamic downloading of CAA records.

[smarnach]

CAA records only apply at the time a certificate is issued, and they only need to be considered by CAs. If the CAA record is changed later, all certificates that have already been issued continue to be valid, even if the new CAA record does not allow the issuing CA anymore. So looking at CAA records would be useless for browsers anyway.

[supriyo-biswas]

Browsers do have this, although this measure is only selectively applied for certain CAs where misissuance has been an issue (There was a Indian CA for which this was used, need to look around MDSP for the link. I’ll post it shortly.)

[agwa]

Historically, root constraints were only used in response to misissuance, but more recently, KamuSM voluntarily limited themselves to .tr when they applied.

[fweimer]

But it doesn't enable covert surveillance. Even without Certificate Transparency, the change in server certificate is visible to the client. Initiatives like Let's Encrypt could make it visible to server operators, too. The browser UI will present those new qualified certificates and existing certificates differently anyway, so I'm not sure if this is going to work.

The bigger issue is that for this in order to work at all, the regulation must have provisions for issuing fake assertions of existing identities to law enforcement and other security services. The predecessor didn't seem to have that. This is different from providing fake identification documents for undercover operations because as far as I understand it, those use are usually mostly made-up and do not impersonate another person.

We would have to read the actual text of the proposed regulation to know the details, but both sides (legislators and those fueling the outrage machine) do not really want us to form our own opinion and hide the draft text from us.

[phasmantistes]

Unfortunately this isn't how it works in practice.

Changes to server certificates happen all the time -- every 60 days or so, if you're getting certs from Let's Encrypt. Browsers can't tell their users every time a certificate changes because the users will just get notification-blindness and be trained to click past the warnings.

Let's Encrypt doesn't help server operators see this; I really not sure what you mean by that. Certificate Transparency would help server operators see this, but the new law text forbids browsers from requiring CT for these certs!

The law doesn't have to solve the problem of how security services will assert fake identities. Each member state can solve that internally. Allegedly, given the recent report of a hijack against jabber.ru and xmpp.ru, they already have. The problem is that, when they do, no one else has any recourse. No other member state can say "hey, don't hijack my websites!", no citizen can say "hey, don't hijack my traffic!", and no browser can say "hey, you issued a false certificate, we don't trust you anymore!".

Fundamentally, the whole issue with eIDAS comes down to one thing: you cannot mandate trust. By definition. If it's mandated, it isn't trust, it's something else. By mandating that browsers "trust" certain CAs, they're breaking the entire trust model of the internet.