Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Encrypted Client Hello

blog.cloudflare.com

Post reply on HN
Print view View original

151–160 of 219 posts

Re: Encrypted Client Hello

#151

Post by YeBanKo »

d-z-m wrote: post #145
YeBanKo wrote: post #72

Earlier quoted context omitted.

> I still think DoH is hot garbage and the way it has been implemented across browsers is an atrocity. Not sure if it's a hot garbage, but I don't see why it's better than DoT or DoQ, except maybe a use case for censored countries. DoT is faster and can be abstracted away from from HTTP. Presumably, DoH is more privacy preserving, because it runs on the same port and looks just like the rest HTTPS traffic. But I thin…

> If it's an HTTPS connection over 443 going to a know DNS server, then it's probably a DNS request, thus I don't see added privacy here. The ISP doesn't see the DNS request, therefore added privacy(you are presumably contacting a DoH server whom you trust). ISPs can pretty much get away with blocking port 853 without much flak(fairly niche, not much use), but if customers' DoH queries to Cloudflare aren't getting ou…

DNS don’t serve regular web traffic, so all the request to DNS server can be classified as DNS request. So what does it add on top of DoT?

Re: Encrypted Client Hello

#153

Post by YeBanKo »

tialaramex wrote: post #79
YeBanKo wrote: post #72

Earlier quoted context omitted.

> I still think DoH is hot garbage and the way it has been implemented across browsers is an atrocity. Not sure if it's a hot garbage, but I don't see why it's better than DoT or DoQ, except maybe a use case for censored countries. DoT is faster and can be abstracted away from from HTTP. Presumably, DoH is more privacy preserving, because it runs on the same port and looks just like the rest HTTPS traffic. But I thin…

Deployability is what matters. DoH had great deployability because everybody speaks HTTPS. In my home, lots of technologies would work. I have static v4 and v6, I have complete control over the firewalls, I can do whatever I want. But at my mum's house, who knows what ports work and which protocols work over them and whether you can change any of that. HTTPS definitely works though, because if it didn't her web brows…

Everybody who speaks HTTPS speaks TLS and the existing DNS-over-TCP. DoT is just DNS-over-TCP tunneled inside TLS. And it can be on any port, not necessarily 853, could be even on 443. What is the point of wrapping it into http request?

Re: Encrypted Client Hello

#154

Post by profmonocle »

yardstick wrote: post #144

> The outer SNI is a common name that, in our case, represents that a user is trying to visit an encrypted website on Cloudflare. We chose cloudflare-ech.com as the SNI that all websites will share on Cloudflare. Because Cloudflare controls that domain we have the appropriate certificates to be able to negotiate a TLS handshake for that server name. All this really means is operators that inspect SNI will now just bl…

If I'm reading this right, blocking this would result in a denial of service rather than a security downgrade. And Cloudflare is so large that blocking all of its hosted sites would effectively look like a partial internet outage. That might not bother state actors, but any ISP in a free country that willingly did this would be so overwhelmed by customer service requests that it wouldn't be worth it.

Any legitimate network operator who wants to avoid this for securiity reasons (i.e. for corporate managed devices) can just disable it by policy, assuming they aren't already using a decrypting TLS proxy.

Plus, there's an easier loophole than blocking ECH - block encrypted DNS. AFAIK most DNS over HTTPS/TLS implementations fall back to plaintext DNS if they can't make an encrypted connection. ECH's reliance on DNS privacy is its weak point; one that can't really be avoided right now.

Re: Encrypted Client Hello

#155

Post by jeroenhd »

assassinator42 wrote: post #41

This is going to make it even more of a pain to do egress filtering on networks/systems we administer. I want to be able to allow list sites with dynamic IPs. The existing solutions for doing this by examining SNI are already often bypassable by forging the SNI (looking at you, AWS Network Firewall).

If you're doing any SNI filtering, you can validate the SNI with the certificate passed in the ServerHello (which most tools skip, because that's computationally expensive in comparison). This involves reconstructing message streams as well, of course. You'll also need to figure out how to deal with 1RTT (probably block it).

I don't think it's much more of a pain than it already was. You can disable the feature on your side if you don't want it, and block+log any traffic with ECH enabled. Hell, you can set up an intercepting proxy with your own CA if you want. That way, nothing gets out without at least getting logged.

Of course this becomes quite difficult if you don't own the devices you're monitoring, but that is kind of the point behind TLS/ODoH/ECH.

Re: Encrypted Client Hello

#156

Post by profmonocle »

elashri wrote: post #114

It is surprising that if you are on the free plan then “Enabled by default for Free zones.”

This is very wise of them. Rolling out new network features is always going to be something of an experiment, no matter how confident you are that the feature meets all the standards and works in all browsers and all OSes. You never know what weird stuff is going to be out there that will screw things up. Back when they first tried to roll out TLS 1.3 there were tons of problems due to buggy middleboxes in corporate networks that couldn't conceive of a new TLS protocol. Thankfully they worked around it, but it probably sucked for the guinea pigs.

CloudFlare's giant footprint of free customers gives them an advantage: it's way more acceptable to roll out new potentially downtime-causing features to users who aren't paying than to businesses paying you hundreds to thousands of dollars per month. The free tier gives them a bit of leeway to experiment, and gives them massive amounts of data from around the world to tell them how those experiments are working.

Re: Encrypted Client Hello

#157

Post by EVa5I7bHFq9mnYK »

ocdtrekkie wrote: post #100
josephcsible wrote: post #50

Earlier quoted context omitted.

> while ECH provides a significant privacy improvement, I personally am against its implementation. Most ISPs enforce country-specific orders to block domains using a combination of DNS packet interception and SNI inspection. The legitimacy or sanity of such laws are a separate matter - countries would want to block websites that violate their laws. That's exactly why I'm in favor of it: it makes effective censorship…

We will almost certainly block ECH at my work, as we already block DoH. I expect any sane network to do the same. The natural alternative if blocking these protocols becomes unsustainable will absolutely be to require full decryption at our security edge. And that will provide drastically more information to us than we have now and will absolutely feel invasive. The idea of uninspectable client traffic is somewhat un…

"Malicious actors" are employees trying to visit instagram from work, right?

Re: Encrypted Client Hello

#158

Post by jeroenhd »

Thoreandan wrote: post #36

Naive question, In 2023 how does DNS-over-HTTP (DoH) affect things like the Cisco Distributed Director / F5 3-DNS where DNS is used to control which datacenters customer traffic is going to? Is this still a technology smaller sites can use?

> where DNS is used to control which datacenters customer traffic is going to

That was never reliable in the first place because of caching and non-compliant DNS servers, but:

EDNS has a DNS protocol extension that will send your IP subnet to the authoritative host even if the request has gone through various resolvers.

Not all DNS servers implement this (notably, Cloudflare doesn't). You can easily check if your DNS server sends your subnet along DNS requests: try to open https://archive.today/. If you see a Cloudflare error or a connection refused/500 error, this EDNS feature is not supported. archive.* intentionally sabotages DNS responses for servers that don't carry these extensions.

BGP anycast would be a solution that can route customer traffic to local datacenters without needing to fall back to DNS hacks and the many broken DNS intermediates.

Re: Encrypted Client Hello

#159

Post by profmonocle »

discardedrefuse wrote: post #15

I feel like this is only possible because Cloudflare is already so huge. If this becomes widely adopted, anyone who wants to offer "private" access to their site will have to move through Cloudflare. This can't be good.

Any CDN or shared hosting provider could implement this same standard. The only way it's truly useless is if a web site has its own IP address / small set of IP addresses.

(Or if the user isn't using some form of DNS privacy like DNS-over-HTTPS/TLS, but then their domain connections are exposed regardless of whether their CDN/host implements ECH.)

Re: Encrypted Client Hello

#160

Post by jeroenhd »

profmonocle wrote: post #154
yardstick wrote: post #144

> The outer SNI is a common name that, in our case, represents that a user is trying to visit an encrypted website on Cloudflare. We chose cloudflare-ech.com as the SNI that all websites will share on Cloudflare. Because Cloudflare controls that domain we have the appropriate certificates to be able to negotiate a TLS handshake for that server name. All this really means is operators that inspect SNI will now just bl…

If I'm reading this right, blocking this would result in a denial of service rather than a security downgrade. And Cloudflare is so large that blocking all of its hosted sites would effectively look like a partial internet outage. That might not bother state actors, but any ISP in a free country that willingly did this would be so overwhelmed by customer service requests that it wouldn't be worth it. Any legitimate n…

In many cases you don't even need to block DoH. Many implementations (Chrome/Edge/I believe Windows itself) will try to use the ISPs DoH servers by default if available, and switch to an alternative if the user chooses to do so.

Because very few websites bother to implement things like DNSSEC (and even fewer clients bother to validate it), the ISP DNS server can then fake all the ECH data it wants,

Post reply on HN