5d ago

DNS Black-holing w/ DNS over TLS - Personal Privacy Part 1

So DNS Black-holing is not new obviously, and what stands out as the go to solution? Pihole probably... and yeah thats what im using because hey its a popular choice. Though I am running it in docker. Combining that with Unbound (also in docker), and configuring outbound DNS to use DNS over TLS, with a few additional minor tweaks, but otherwise mostly standard configuration on both.

Wondering what you guys might be using, and if you are using Pihole and/or Unbound if you have any tips on configuration.

Happy to share my config if there is interest.

18 comments

I got two PiHoles running on my network via Docker Compose, I tried setting up Unbound in Docker-Compose and that fell flat, from my understanding DNSSEC was preventing DNS resolution outright.

Also tried OpenSense + Unbound which led to the same thing.

Eventually got tired of having my network cutting in and out over minor changes so I just stuck with Quad9 for my upstream needs.

happy to share my docker-compose with pihole and unbound. im not the original author its a compilation of a few peoples. no issues. normal DNS inside the house DoT outside.

If you don't mind DM'ing me or dropping it in a comment here it would be greatly appreciated! The docker engine isn't something entirely new to me so i'm a bit skeptical into thinking that i missed something but always happy to compare with others, actually Docker is what pushed me to switch fully to Linux on my personal computers.

Snippet from my docker-compose.yml:

    
  pihole:
    container_name: pihole
    hostname: pihole
    image: pihole/pihole:latest
    networks:
      main:
        ipv4_address: 172.18.0.25
    # For DHCP it is recommended to remove these ports and instead add: network_mode: "host"
    ports:
      - "53:53/tcp"
      - "53:53/udp"
      - "127.0.0.1:67:67/udp" # Only required if you are using Pi-hole as your DHCP server
      - "127.0.0.1:85:80/tcp"
      - "127.0.0.1:7643:443"
    environment:
      TZ: 'America/Vancouver'
      FTLCONF_webserver_api_password: 'insert-password-here'
      FTLCONF_dns_listeningMode: 'all'
    # Volumes store your data between container upgrades
    volumes:
      - './config/pihole/etc-pihole:/etc/pihole'
      - './config/pihole/etc-dnsmasq.d:/etc/dnsmasq.d'
      - '/etc/hosts:/etc/hosts:ro'
  #   https://github.com/pi-hole/docker-pi-hole#note-on-capabilities
    cap_add:
      - NET_ADMIN # Required if you are using Pi-hole as your DHCP server, else not needed
      - CAP_SYS_TIME
      - CAP_SYS_NICE
      - CAP_CHOWN
      - CAP_NET_BIND_SERVICE
      - CAP_NET_RAW
      - CAP_NET_ADMIN
    restart: unless-stopped
    labels:
      - "traefik.enable=true"
      - "traefik.http.routers.pihole.rule=Host(`pihole.my.domain`)"
      - "traefik.http.routers.pihole.entrypoints=https"
      - "traefik.http.routers.pihole.tls=true"
      - "traefik.http.services.pihole.loadbalancer.server.port=80"
      - "traefik.http.routers.pihole.middlewares=fail2ban@file"
 unbound:
   image: alpinelinux/unbound
   container_name: unbound
   hostname: unbound
    networks:
      main:
        ipv4_address: 172.18.0.26
   ports:
     - "127.0.0.1:5334:5335"
   volumes:
     - ./config/unbound/:/var/lib/unbound/
     - ./config/unbound/unbound.conf:/etc/unbound/unbound.conf
     - ./config/unbound/unbound.conf.d/:/etc/unbound/unbound.conf.d/
     - ./config/unbound/log/unbound.log:/var/log/unbound/unbound.log
   restart: unless-stopped

Edit: After re-reading the Unbound github and their documentation it seems i may have missed some volume mounts that are key to the function of Unbound, i'll definitely have to dive deeper into it.

Out of curiosity, why did you decide on DoT instead of DoH (DNS over HTTPS)?
- totally arbitrary, lol. Im used to DNSSEC, saw DoT and DoH about the same time, think I saw a write up that used DoT and just went for it. Havent even compared DoT vs DoH, but DoH reminds me of Homer Simpson cuz im old XD
  
  Dammit i read it as DOH in a homer simpson voice in my head aswell!!
I have an n100 box that I put opnsense on for routing, firewall, DHCP, DNS and IDS. It uses unbound for DNS and so I'm leveraging the blocklist functionality in unbound. And then I use unbound to resolve instead of using DoT forwarding.
Dnsbl is only a small component of effective network security. Arguably the firewall is most important and so I have a default deny all for any device on my LAN trying to reach the Internet.
All applications need specific allows. Thus internally no device can use dns over tls because 853 is blocked by default. Then I use a DNSBL to catch known DoH by domain since the cert is provided by domain name.
- "Dnsbl is only a small component of effective network security. Arguably the firewall is most important and so I have a default deny all for any device on my LAN trying to reach the Internet." 100%, I decided to break up my posts into sub components of the total stack, but to your point currently im enforcing a deny all inbound and outbound at the host level, as the network is shared with the fam and they are not ready for that level of learning (pain, lol)
  I just learned about unbound, didnt realize it had a blocklist capability so thats great to know. Gotta dig into it.
  I like that last bit, blocking DoT except for the one approved path. Much like TLS 1.3 it offers insider threat protection against inspection. So with that in mind when you said you are using unbound instead of using DoT forwarding, you mean instead of allowing clients to DoT forward, right? Thats what I am doing now as well, though I am not actively blocking it yet. Just currently enabling and testing feasibility on a single host to see the performance and operational impacts of privacy/security implementations.
  Curious to your IDS solution, I gotta dig into opnsense. I know about it, its been around a long time, but havent touched it in so long I cant remember its capabilities.
  
  Wrt lan deny all for the fam, it's mostly hard on gamers cuz games tend to use wide port ranges and outbound IPs are potentially home isp networks not the game servers. But yeah it takes some time and research to really lock it down.
  Most stuff is running through web protocols though. So right off the bat you create allow rules for any LAN device to hit ports: 80, 8080, 443, 8443 which are your common http and https ports. That's gonna get most ppl what they need.
  I do ASN based allows for certain applications like Google, Facebook, etc.
  For consoles they're pretty locked down so just give them full allow to the Internet. I don't do that actually but it's probably the better way.
  IOT devices get only the ports they need to the IPs they need.
  when you said you are using unbound instead of using DoT forwarding, you mean instead of allowing clients to DoT forward, right?
  No I mean my unbound resolves DNS for something like microsoft.com all by itself. It calls up the root name servers, finds the com nameservers, then asks the com nameservers for Microsoft. And for any subdomains it asks the MS name servers. This is instead of relying on external forwarding services like 8.8.8.8 or 1.1.1.1 or quad 9 or whatever. At least the former two are sure to be aggregating this data.
  Additionally I do not allow devices on my network to reach out to external port 53, or 853 to circumvent lookups on my unbound by reaching out directly, which would then bypass the DNSBL. Anything for port 53 gets NAT'd to the unbound server. You can't redirect TLS attempts so those get hard blocked.
  Curious to your IDS solution
  Securicata is what opnsense uses. Pretty easy to set up.
Unbound on the router which connects upstream with DNS over TLS. Ports 53 and 853 are NATed to the phiole and several other DNS servers like Google's are blocked so devices can't bypass the pihole very easily. This is only on my primary VLAN. Other VLANs are given the Unbound DNS by default but are allowed to bypass if they insist. I have one VLAN for guests and one for trusted devices in addition to the primary one.
- In my particular setup, I have an additional constraint and that is that my network has to be designed for portability and travel. Not that it affects your design per say. Thank you for the response. Just something that occurred to me that I hadnt mentioned.
  I am living a transient life at the moment. So lots of virtualization and lack of control concerning the WAP and such.
  I do like your set up btw.
I'm using a service called Control D, which has 4 levels of free DNS. One is no blocking at all. One is only blocking malware. One is blocking malware and ads, and the other one is blocking malware, ads, and social media, like Facebook.
I use cloudflared to translate DNS into DNS over TLS instead of Unbound to make it into recursive DNS. Just never really seen the need to switch it. I'm happy with nextDNS + Cloudflare resolving DNS upstream.
The main thing I wanted to note is port 53 outbound is blocked at the router to prevent devices from using external/unencrypted DNS. If a LAN device wants DNS resolution they MUST use the LAN DNS servers they were given via DHCP, or use their own DoT config, as plain DNS won't make it out of the network.
It's because of this block/enforcement that I run two local DNS servers: pihole on an RPI and a mirror on my main server tower, with Galaxy-Sync keeping them identical. If I tinker with/update one, the other picks up the slack so connectivity/resolution isn't disrupted.
- Yeah, I am pretty close to that, the pihole to unbound, unbound DoT to cloudflare. What I am doing at this point is bypassing the DNS to ISP, but as I stated in my response above, not yet blocking everything on the net from using the regular stuff. Just feasibility testing at the moment.
  Love the dual setup for DNS. I set my primary to this and my secondary to just cloudflare at them moment for when I bork my primary DNS will fidgeting with it, haha.

18 comments