4 ms·
It would be nice if Firefox supported some way to automatically rotate which backend provider it's using for DoH. This would keep any one entity from having a
by tyingq 5y ago
It would be nice if Firefox supported some way to automatically rotate which backend provider it's using for DoH. This would keep any one entity from having a continuous feed of every DNS lookup.
- moltar 5y agoGreat idea. Would a local load balancing proxy do the trick?
- vngzs 5y agoNot with a TCP proxy. To establish a HTTPS tunnel, the client needs to know the identity of the remote end for certificate validation. Randomizing remote ends with a naive proxy would lead to TLS errors. So you'd likely configure your browser to do DoH destined for that proxy, and have the proxy generate DoH requests to the remote ends. The downside is you'd have to somewhat manually set up certificate trust between your client and your proxy - either by trusting the proxy's certificate, or installing a root CA that issues the proxy's certificate in your browser's trust store.
- jeroenhd 5y agoI use my PiHole to do balancing like this. I've set it up with DoT and DoH so that DNS requests are spread across several servers. Then, I expose my PiHole as a DoT + DoH server (DoT for Linux and Android, DoH for Firefox) so that I can always use my ad block lists and secured DNS traffic.
- Thiez 5y agoDoesn't this just result in more organizations having your data? Presumably most DNS entries have some finite TTL, such as a few minutes / hours / days. Let's say one of your favorite websites has a TTL of 10 hours, and you visit this website once per day. Then eventually (soon!) you will have asked most or all of your servers for that domain, so now they all have your data. Or do you have your PiHole configured that it always asks the same server for the same domain?
- 3np 5y agoYou could share the namespaces using some consistent hashing so any given domain would hit the same server. Maybe something exists for that already?
- jeroenhd 5y agoYou're right that this is the case. I set it up so that I can use the double encryption system with a DoH proxy when I eventually resolve this issue. In practice, 60% is my traffic ends up at cloudflare already, so I might as well use their DNS until then. It's also an availability thing. There's that one archive website on a vendetta against cloudflare for not forwarding enough data in DNS requests (eDNS or something?) that I can access by using alternative servers as well. In the rare case of a cloudflare outage, this also ensures uptime. DNS cannot be made fully anonymous, though that double encrypted DoH thing seems promising. Picking a few trusted providers isn't a problem for me personally.
- unethical_ban 5y agoI don't understand. I have a PiHole, and have the cloudflare DoH client installed on it. Requests from my network go to the PiHole, and the PiHole is set to query localhost:5053, which then hits a DoH provider. Why do you have separate setups for mobile vs. desktop web browser, why not have the OS do native DNS to your PiHole?
- jeroenhd 5y agoWhen I'm on my laptop, in the train, working on Windows, I can't use my home DNS without exposing it to the world and becoming part of every DDoS attack there is. So, for those situations, I've got Firefox set up to do DoH. Windows can do some DNS encryption through a complex setup of local DNS servers but it's nothing as easy as the standard Linux setup. Similarly, I use DoT on my phone but if I want to connect to any hotspot, I have to disable DoT because the hotspot login detection fails with DoT configured. I think it's a Xiaomi thing. Regardless, I need to disable it to use. In one particular case (scanning groceries in the supermarket for the automated payout) I can't fall back to LTE like usual. Just in case I forget to turn DoT back on, I've also configured Firefox on my phone to use DoH because the mechanism is there and I might as well. There's small edge cases where I just like to fall back to browser DoH. DoT and I'm network DNS still works fine, of course.
- tristor 5y agoThat's an interesting idea. I wonder how you'd feel about this in relation to the trade-off against multiple resolver providers having information about your Internet traffic? By using a single provider, you limit your risk exposure, but also increase your reliance on that provider. So it's not clear-cut that this would necessarily be beneficial.
- tyingq 5y agoThat's a fair point. I think I would still be happier if no one entity had everything. But I do recognize the downside.
- rakoo 5y agoIt's actually a very good point, and Tor is explicitely using a single entry guard because of that: it was analyzed that keeping a single guard for a long time is "on average" better than regularly changing guards.
- goodpoint 5y agoRotating between a handful of providers could be a net negative for privacy: - every FQDNs you resolve is leaked to at least one provider - the FQDNs you resolve often (e.g. every day) are leaked to at all the backend providers over time If you want to protect your privacy you need DNS resolution over Tor.
- JoshTriplett 5y agoWhile I agree that rotating might be a net negative, that particular issue could be worked around: hash the FQDN and use the hash to determine which provider you use, so that a given user always resolves a given FQDN with the same provider. (Salt the hash, so that different users use different providers for that FQDN.) It's still not clear if it's a win to rotate, but if you were going to rotate, that would help.
- goodpoint 5y agoThat's not rotation, by definition. (and hardly beneficial for privacy: you just divided the FQDNs you leak across 3-4 providers)
- JoshTriplett 5y agoThe only advantage would be in reducing the ability to correlate a user's sites. But nonetheless, I agree.
- np-tokumei 5y agoDistributing DNS queries across several resolvers is a good idea to make sure there's no single resolver would be able to learn the entire browsing history of a user. However, this distribution algorithm should be designed with care because distributing queries in a random or round-robin manner would lead to even worse privacy since every resolver will eventually learn the entire browsing history of the user. We have conducted some preliminary work on this idea and shared it with the team at Firefox, but I guess the current number of Trusted Recursive Resolvers is not large enough. However, I do believe that once Firefox have more than a dozen TRRs, (or if Firefox allows users to input more than just one customized DoH resolver), our proposed K-resolver resolution mechanism would help to increase the privacy benefit of DoH even further. "K-resolver: Towards Decentralizing Encrypted DNS Resolution" https://www.ndss-symposium.org/wp-content/uploads/2020/02/23009-paper.pdf https://www.ndss-symposium.org/wp-content/uploads/2020/02/23...