When we audit a corporate network, the first thing we check is what happens to DNS on staff laptops. In plenty of offices the answer is unsettling. Devices quietly send every lookup to public resolvers, and the security team has no idea it is happening. Encrypted DNS closes the eavesdropping gap, but it can also switch off the visibility your monitoring depends on. So when we weigh DNS over HTTPS vs DNS over TLS for enterprises, the real question is not which protocol is newer. It is who is allowed to see the queries. Our team at MONA Media helps us keep technical guidance like this accurate, but the testing and opinions here come from our own lab work.
Our short answer: for most organisations, DNS over TLS (DoT) or a managed internal DoH resolver is the safer default. DoT wraps DNS in TLS on TCP port 853, so your network controls can still recognise the traffic as DNS. DoH (RFC 8484) wraps DNS in HTTPS on port 443, mixed in with ordinary web browsing. That makes it far harder to spot. Once browsers are free to pick arbitrary public DoH endpoints, your grip on name resolution slips away quickly.
The real trade-off sits between privacy and visibility
Encryption is a genuine improvement. A plain-text query can be read, and even rewritten, by anyone on the path. That might be the operator of a hotel Wi-Fi network or an attacker sitting on a compromised access point. Encrypting the query protects laptops that spend time on networks you do not control.
The catch is that the same encryption hides those queries from you. Your DNS firewall, your response policy zone feeds and your detection sensors all depend on seeing the name being requested. Once that traffic is encrypted and leaves by another path, those tools go quiet. Nothing alerts you when it happens, and that silence is what makes it dangerous.
Settle one question before you pick a protocol: who should see the queries? It could be your own resolver, a managed third party, or nobody at all. That decision shapes everything that follows.
How the two protocols look on the wire
DoT runs over TCP port 853 with TLS. DoH runs HTTP/2 over TLS on TCP port 443 and uses the application/dns-message media type. Because DoT owns its own port, a firewall can identify it by port number alone. DoH was designed to blend into HTTPS, so spotting it takes SNI inspection, certificate checks or endpoint-based controls.
The RFCs define the semantics, but individual products choose their own defaults. Confirm the actual server behaviour in your resolver vendor’s documentation before you assume it matches the specification.
What breaks in your security tooling
DNS firewalls and sinkholes lose sight of queries the moment clients bypass the internal resolver over DoH. A sinkhole still protects the clients that use it, but a laptop talking to a public DoH endpoint simply walks around it.
Detections for DNS tunnelling, domain generation algorithms and fast-flux lose their input in the same way, unless the traffic is forced through the monitored resolver. In our own lab runs, the detections that mattered most were the first to stop firing. That is exactly why we retest them after every change instead of trusting that they still work.
Endpoint and proxy telemetry can fill part of the gap, but it does not replace resolver logs. Support for DoT and DoH on firewalls and proxies also varies between software releases, so check current vendor notes before you assume a feature exists on your gear.
Locking clients to your own resolver
Start at the perimeter. Block outbound TCP and UDP port 853 except from your internal resolver’s address. That closes the easy route for DoT and for any client that tries it.
DoH is harder, because it hides inside port 443. Block the known public DoH endpoints by domain and IP, and restrict DoH on port 443 wherever your platform allows it. Be careful with allowlists. Breaking legitimate browser or app behaviour is the most common complaint we hear during rollout, so test each exception before it goes live.
Push a resolver policy to endpoints through group policy or MDM, pinning the internal DoH or DoT server. Endpoint policy on its own enforces nothing against a determined user, so always pair it with network-level blocking. On an internal BIND resolver that forwards upstream over DoT, the forwarding block looks roughly like this:
- forward only;
- forwarders { 192.0.2.10 port 853 tls …; };
Check the exact tls and forwarder syntax against your BIND version’s documentation before you deploy it.
Checks you can run today
First, verify which path a client actually takes. Run dig +tls @resolver example.com to test DoT, then run a DoH query with curl against your internal DoH URL. Compare the results with what the client does by default, because the gap between the two is often where the problem hides.
To confirm the port blocks, run a controlled test from a guest VLAN. A blocked TCP 853 connection should time out. If it connects, the perimeter rule is not doing its job, and you need to find out why before going further. Finally, confirm that the internal resolver still logs every query after the change. A rollout that quietly drops logging has solved the wrong problem.
Setting the two side by side
DoT is usually the simpler option to govern. It sits on a dedicated port, so one rule covers it, and it produces logs your SOC already knows how to read. DoH demands more effort because it travels in the same traffic as ordinary browsing, so you need SNI or endpoint controls to tell it apart. Neither protocol is safer in itself. The difference lies in how easily you can keep the resolution path inside your control.
A typical rollout in a Vietnamese office
Picture an office of 200 to 500 seats, with one internal resolver and a next-generation firewall doing DNS inspection. The design has to sit alongside local ISP resolvers and cloud DNS paths, which often behave differently from the lab. We have seen a clean design fail in production because one branch quietly used its ISP resolver for half its traffic.
Run the rollout as a pilot group first. Measure resolution latency and log completeness before enforcing anything across the whole office. Any figures you quote for your own network should come from your own measurements, not from generic claims you read elsewhere.
Common mistakes when rolling out encrypted DNS
The most frequent mistake is blocking DoH on port 443 without an allowlist. That breaks legitimate browser and app behaviour and fills the helpdesk queue within an hour. The second is forgetting mobile and BYOD devices, which often ignore the managed resolver setting entirely. The third is assuming endpoint policy enforces the resolver by itself, and it does not.
Frequently asked questions
Can I just disable DoH in the browser?
Often yes, through browser policy, but that only covers managed browsers. A user running another app, or an unmanaged device, can still send DoH traffic. Pair browser policy with network-level blocking.
Does DoT still leak metadata about queries?
DoT encrypts the query content, but the connection still reveals that a device is talking to a resolver, along with the timing and volume of its lookups. Encryption protects the name being looked up, not the fact that a lookup took place.
Should I run my own DoH or DoT resolver?
If you have the staff to operate it, running your own resolver keeps the logs in your hands. If you do not, choose a managed service you can audit. Either way, keep name resolution inside your control.
How do I keep SOC detections working after the change?
Force all DNS through the monitored resolver, confirm the logs still arrive, and retest your tunnelling and DGA detections. Do this with the pilot group before the full rollout, not afterwards.
Do firewalls inspect DoH traffic?
Some can, through SNI or certificate inspection, but not all of them. Check the current release notes for your firewall, since support varies by version and you should not assume the feature exists.
If you are rolling this out now, start with a pilot, measure what you can see before and after, and make the internal resolver the only path clients are allowed to use. For more on controlling traffic between services, read our piece on mTLS in hybrid cloud, and for the wider security picture see our cybersecurity coverage. Whatever you decide, write down who is allowed to see DNS queries, because that answer will still matter long after the protocol itself has changed.