Hacker News Reader

Comment by ga-vu | Hacker News Reader

My problem is that this is highly overhyped by the Mozilla/Cloudflare PR teams, and both companies stand to make money out of it.

A better solution would be DoT+DNSSEC: https://twitter.com/jschauma/status/1184483451111727106

I don't trust Mozilla anymore, especially after becoming a VPN vendor and partnering with Cloudflare. They now have commercial interests in pushing standards down on everybody, similar to Google.

View on HN

My ISP (Entel Chile) started blocking DNS requests to non-Entel servers in July. DoT would not give me a way around this, because they can still see that it is DNS traffic by the port. DoH mixes DNS traffic in with the noise of HTTPS traffic, allowing me to run my own DoH resolvers and bypass their restrictions.

Once your ISP starts mucking with and blocking your traffic, what you need is a VPN and/or Tor – DoH is a very partial solution at best. What DoH gives you is merely a correct answer to a DNS lookup of an IP address. But if your ISP blocks traffic to that IP address, which they could, you’re still blocked; DoH did not help you.

The ISP is trying to block other DNS traffic, they're not necessarily trying to block certain websites (or if they are, they wouldn't want to block [for example] all of Cloudflare or the GCP load balancers to block porn hosted on either of those services). DoH solves that problem as directly as possible without going to solutions that have other side-effects.

The ISP do most certainly want to block web sites. The currently most common method is to block DNS, but they could, and will, switch to IP blocking in a heartbeat if they have to.

> they wouldn't want to block [for example] all of Cloudflare or the GCP load balancers

If your only defense against blocking by your ISP is to centralize web server hosting, then you’re just moving the problem to there instead. Cloudflare also blocks, or “deplatforms”, those they deem unseemly. And once someone is taken off the centralized web server hosts, they can be blocked by your ISP again. You’re not solving the problem in the long term.

DoH is fighting yesterday’s war.

Tor has limited value. Lots of things are blocked (search, for example), and it's slower than dialup. The security of the exit points is also questionable, and this is about trust.

Pushing all traffic through a VPN only moves the problem. Now I have to trust some unknown entity in some other country to respect my privacy. I don't want my traffic to bounce from Chile to Norway and then back to Chile when I do online banking. I run a VPN endpoint in Chile for my own traffic, but I don't need to push all of my home traffic through it....yet.

All useful traffic runs over HTTPS already, but metadata is still valuable. My ISP could snag SNI headers and know sites, but I don't think they're actively sniffing 100% of the traffic that goes across the wire. No, I think they want to be Verisign/Comcast/OpenDNS and redirect NXDOMAIN responses to ad pages or inject false responses to "unauthorized" queries.

> Tor [is] slower than dialup.

That was true years ago, but nowadays I rarely percieve any speed difference at all when comparing browsing using Tor and non-Tor.

> Pushing all traffic through a VPN only moves the problem. Now I have to trust some unknown entity in some other country to respect my privacy.

But, and here’s the crucial difference: with VPNs you have a choice. A wide variety of choice. With ISPs, especially in the US, not so much.

> All useful traffic runs over HTTPS already, but metadata is still valuable. My ISP could snag SNI headers and know sites,

That’s due to be fixed with ESNI. Just wait for a technical fix.

> but I don't think they're actively sniffing 100% of the traffic that goes across the wire. No, I think they want to be Verisign/Comcast/OpenDNS and redirect NXDOMAIN responses to ad pages or inject false responses to "unauthorized" queries.

That is true today, but when DoH and/or DoT happens, the ISPs will certainly switch to doing whatever still works. They are merely sniffing and proxying DNS traffic today because it still works, but the second it no longer works, they will switch to whatever does work. You can’t just work around the current mechanisms, you have to look a few more moves ahead.

Mozilla has deployed a canary domain with Firefox. Your ISP can block that canary domain and they'll automatically disable DoH for you... ooops!

> DoH mixes DNS traffic in with the noise of HTTPS traffic, allowing me to run my own DoH resolvers and bypass their restrictions.

And when malware gets on your network and starts using DoH to get around your home resolvers, what will you do?

* https://www.zdnet.com/article/first-ever-malware-strain-spot...

* https://www.zdnet.com/article/psixbot-malware-upgraded-with-...

As someone who works in IT, this is my problem with DoH.

That was already a possibility even before all of this DoH publicity. Mozilla, etc. pushing DoH publicizes it's availability, but there was nothing in the past preventing malware from tunneling all sorts of traffic over HTTPS. DNS inspection isn't an end all, be all for malware security. It just gets the low hanging fruit.

There was a lot of low-hanging fruit given that most malware writers aren't going to set up all of this infrastructure for custom protocols.

And even when they did, creating various C&C servers, the lack of ESNI would allow for detecting activity once the daily domain creation algorithm was reverse-engineered:

* https://blog.malwarebytes.com/security-world/2016/12/explain...

This is possible regardless of whether Chrome/Firefox/etc deploy DoH support. In fact, it was possible before DoH at all: nothing mandated that malware use standard DNS requests on port 53.

> ... nothing mandated that malware use standard DNS requests on port 53.

Except using other ports would cause network monitoring software to throw red flags because of the "strange" traffic on non-standard ports.

Nothing requires malware to pick a “strange” port. It could just shove its data into ICMP, or make it look like an SSH connection on 22, or anything else.

Honestly, I don't see why people keep bringing this up. Malware has millions of ways of creating covert channels and you will never be able to block/detect all of them. It's also well within the capability of major DoH server operators to detect and shut down malware domains pretty fast.

If you're going to make these sorts of claims you need to also lay out how you logically came to them. What is the financial path incentive for Mozilla since they already have the ability access to anything about every page you ever go to? On the Cloudflare side have you looked into the resolver agreement and based your statement around those limitations or just your own conjecture https://developers.cloudflare.com/1.1.1.1/commitment-to-priv... ?

DNSSEC doesn't provide privacy at all; it signs records, but doesn't encrypt queries or responses. So it's orthogonal to the DoH debate.

DoT does essentially the same thing that DoH does. The difference between DoT and DoH, from a practical perspective, is that DoT uses a new TCP port, and DoH overloads 443. The reason DoH overloads 443 is to make it difficult for network operators to disable. In other words: DoT is simply DoH with a network-controlled kill switch.

How exactly is DoT "more comprehensive" (to use the words from the tweet you've cited) than DoH?

Reply on news.ycombinator.com