Skip to content
Shield

Explainer

Protective DNS vs DNS filtering vs DNS firewall: the terms vendors blur

You will be able to tell protective DNS, DNS filtering and a DNS firewall apart, see where each acts on guest and staff WiFi, and know what DNS-layer blocking cannot see. You can then test any vendor on four points - new-domain coverage, per-VLAN policy, failure behaviour and bypass handling - whatever the product is...

12 min read

Protective DNS vs DNS filtering vs DNS firewall: the terms vendors blur

Watch the explainer

Protective DNS is a recursive resolver service that checks each domain lookup against threat intelligence. It refuses to resolve malicious domains, so no connection ever starts. DNS filtering is the wider policy practice. A DNS firewall is the enforcement mechanism, often built on Response Policy Zones (RPZ). The UK NCSC and US CISA run national protective DNS services.

What protective DNS, DNS filtering and a DNS firewall each mean

All three terms describe decisions made at DNS resolution. Resolution is the moment a device asks a resolver to turn a name such as example.com into an IP address. RFC 1034 and RFC 1035 define the process. The lookup happens before any TCP or QUIC connection opens. A decision taken here stops traffic before a firewall or proxy sees a single packet.

The three terms differ in what they name:

  • Protective DNS (PDNS) names a security service. A recursive resolver checks each query against threat intelligence on malware, phishing, command-and-control and newly registered domains. It refuses to resolve the bad ones. The term comes from government. The UK NCSC runs a PDNS service for public sector bodies. The US CISA offers a protective DNS resolver to federal civilian agencies.
  • DNS filtering names a policy practice. It allows or blocks lookups by rule, covering security categories and acceptable-use categories such as adult content or gambling. Protective DNS is DNS filtering with a security mandate. Content filtering on guest WiFi is DNS filtering with an acceptable-use mandate.
  • A DNS firewall names the enforcement mechanism inside the resolver. Rules match a query or response and rewrite it. The rewrite might be NXDOMAIN (domain does not exist), a block page address or a sinkhole, which is a controlled address that absorbs the traffic. The common open format is RPZ, specified in an IETF Internet-Draft and supported in BIND, Unbound and PowerDNS Recursor.

Put plainly, protective DNS is the service and its goal. DNS filtering is the policy, and the DNS firewall is the rule engine. One product can be all three at once, which is why vendor pages blur them.

How does protective DNS work at the point of resolution?

A device on your WiFi does not pick its resolver at random. When it joins, DHCP hands it resolver addresses through option 6, defined in RFC 2132. Point that option at a protective resolver and every device that follows network settings sends its lookups there. That covers guest phones, staff laptops, smart TVs and payment terminals, with no agent installed on any of them.

The resolver then runs four steps for each query:

  1. Identify the network. It maps the query's source to a network or VLAN (virtual LAN, a logically separate network on shared hardware) and loads that policy.
  2. Check the policy. It compares the domain against threat feeds, domain-age signals and content categories.
  3. Answer or refuse. A clean domain resolves as normal. A blocked domain gets NXDOMAIN, REFUSED or a block page address. RFC 8914 Extended DNS Errors lets the resolver say why, with code 15 (Blocked) or 17 (Filtered).
  4. Log the verdict. Every query and decision is recorded. You see what devices on each network tried to reach.
device needs an IPDNS queryapply network policyclean domainmalicious or out of policydevice connectsnothing to connect toevery decisionDevice joinsguest WiFiDHCP option 6points it at theprotectiveresolverLookup forshop-login.exampleStub resolversends the queryon port 53ProtectiveresolverMaps the queryto its networkor VLAN andloads thatpolicyPolicy checkThreatintelligence,domain age andcontentcategoriesAllowed:real IPreturnedResolverrecurses asnormal andcaches theanswerBlocked:block pageor NXDOMAINExtended DNSError 15(Blocked) or 17(Filtered), RFC8914VerdictloggedNetwork, domain,category anddecision forreportingConnectionopensFirewall andproxy rulesstill apply tothe sessionNoconnectionattemptedNo TCP or QUIChandshakereaches themalicious host
How protective DNS decides at the point of resolution, before any connection opens

Flow showing a device on guest WiFi sending a lookup to a protective resolver, which checks policy and either returns the real IP so the device connects, or returns a block response so no connection is attempted, with every verdict logged.

Connections

  1. Device joins guest WiFi to Lookup for shop-login.example (device needs an IP)
  2. Lookup for shop-login.example to Protective resolver (DNS query)
  3. Protective resolver to Policy check (apply network policy)
  4. Policy check to Allowed: real IP returned (clean domain)
  5. Policy check to Blocked: block page or NXDOMAIN (malicious or out of policy)
  6. Allowed: real IP returned to Connection opens (device connects)
  7. Blocked: block page or NXDOMAIN to No connection attempted (nothing to connect to)
  8. Policy check to Verdict logged (every decision)

Why placement before the connection matters

A network firewall decides on IP addresses and ports. By the time it acts, the device already has an answer and is opening a session. Modern hosting weakens IP-based decisions. Many domains share content delivery network addresses, and attackers rotate IPs to stay ahead of IP blocklists. A domain-level verdict at lookup time stops the phishing click or malware callback before the handshake.

It is also light. A resolver answers one small query rather than inspecting a full traffic stream. NSA and CISA set out the same reasoning in their joint guidance, Selecting a Protective DNS Service.

Protective DNS vs DNS filtering vs DNS firewall: where the boundaries sit

The table separates the three terms from four technologies they are often confused with. Read across a row to see where each one acts and what it does not do.

Protective DNS vs DNS filtering vs DNS firewall, and the terms they get confused with Data as of 2026-06-10
TermWhat it isWhere it actsGoalWhat it is not
Protective DNSA resolver service that refuses to resolve malware, phishing and command-and-control domainsRecursive resolver, before any connectionSecurityA content-only filter, or DNSSEC
DNS filteringPolicy that allows or blocks lookups by threat or content categoryResolver, per network, VLAN or deviceSecurity and acceptable usePage or URL-level filtering
DNS firewallRule engine that rewrites matching queries or responses, often via RPZInside the resolver softwareEnforcementA packet-inspecting appliance
DNSSECSignatures proving a DNS answer is authentic (RFC 4033)Authoritative servers to validating resolverIntegrityA policy block of any kind
Encrypted DNSDNS over TLS (RFC 7858) or over HTTPS (RFC 8484)Device to resolverPrivacy in transitFiltering; it can route around a filter
Network firewallAllows or denies traffic by IP, port and protocolNetwork edge, after the lookupAccess controlDomain-aware; it sees IPs only
Secure web gatewayProxy that inspects URLs and contentTraffic path, after connectionContent inspectionResolver-based; traffic must be steered to it
DNS is defined in RFC 1034 and RFC 1035. Extended DNS Errors (RFC 8914) let a filtering resolver report why it blocked. RPZ is specified in an IETF Internet-Draft, not an RFC.
Secure web gatewaySees URLs and content after the connection opensNetwork firewallAllows or denies IPs and ports once the lookup hasreturnedProtective DNS: the serviceSecurity goal: refuse to resolve malicious domainsDNS filtering: the policyAllow and block rules by threat and content categoryDNS firewall: the mechanismRule engine in the resolver, often Response PolicyZonesDNSSEC and encrypted DNSIntegrity (RFC 4033) and privacy (RFC 7858, RFC 8484),not policy
Where each term sits: the three DNS terms share one layer and name different parts of it

Layered diagram with secure web gateway and network firewall above the DNS resolution layer, where protective DNS is the service, DNS filtering is the policy and a DNS firewall is the mechanism, with DNSSEC and encrypted DNS beneath as integrity and transport.

Connections

  1. Protective DNS: the service to Network firewall (decides first, before connect)
  2. Network firewall to Secure web gateway (decides next, after connect)
  3. DNS filtering: the policy to Protective DNS: the service (defines what the service blocks)
  4. DNS firewall: the mechanism to DNS filtering: the policy (enforces the policy)
  5. DNSSEC and encrypted DNS to DNS firewall: the mechanism (secures and carries the answer)

Where people get it wrong

Five errors come up repeatedly when teams compare DNS security products.

  • "DNS filtering is just content filtering." Blocking gambling sites is one use. The same mechanism blocks phishing and malware domains. A content-only configuration leaves the security categories switched off.
  • "A DNS firewall is a box." It is a set of rules in resolver software, run on a server, a firewall's DNS proxy or a cloud service. "Firewall" describes the behaviour, not the form factor.
  • "We have DNSSEC, so we have protective DNS." DNSSEC lets a resolver trust that an answer is genuine. It will validate a correctly signed phishing domain without complaint.
  • "Encrypted DNS protects us." DoT (DNS over TLS) and DoH (DNS over HTTPS) stop eavesdropping in transit. They also let a browser send lookups to a resolver you did not choose.
  • "Protective DNS replaces the firewall." It decides earlier, not instead. Traffic to raw IP addresses never touches DNS.

How vendors label the same capability

Cisco Umbrella markets resolver-based blocking as DNS-layer security, the entry point to a wider cloud security stack. It runs cloud resolvers that apply policy before a connection opens, which is the same placement Purple Shield uses. Ignore the label and test four things. Check threat coverage on new domains, policy per network and VLAN, behaviour on failure and bypass handling.

Where do you meet protective DNS on guest and staff WiFi?

Wherever you give network access to devices you do not manage, the resolver is the one control point every device touches.

  • Hotels and pubs. Guest WiFi needs acceptable-use filtering so your brand is never attached to illegal content. Malware blocking also protects the guest's own device.
  • Retail. Shopper WiFi sits beside staff and point-of-sale networks. Separate policies keep the shopper network open and the payment network locked down.
  • Stadiums and events. Thousands of fan devices join within minutes. A resolver policy needs no per-device setup and no inline appliance.
  • Healthcare. Patient and visitor WiFi shares a site with clinical systems, so malicious lookups must stop before they reach anything.
  • Workplaces. Staff WiFi is where phishing lands. Blocking the credential-harvesting domain removes the page even after the email gets through.
  • Multi-tenant buildings. Residents expect home-style freedom, so policy usually stays security-only.

A per-VLAN policy in practice

A typical venue runs three policies on one estate:

  • Guest VLAN: security categories plus adult, gambling and piracy blocks, with a block page that explains why.
  • Staff VLAN: security categories, newly registered domain blocking and phishing protection, with fewer content blocks.
  • Payment and IoT VLAN: security categories plus a tight allowlist of the domains the devices need. This narrows what a compromised device can reach on a network you already segment to reduce PCI DSS scope.

The resolver knows which VLAN each query came from. The same domain can get three different answers.

What protective DNS cannot see: limits and edge cases

Protective DNS is an early, inexpensive decision. It is not a complete one. These are the edges to design around.

Devices that skip your resolver

Some apps hard-code public resolvers. Browsers can send lookups over DoH (RFC 8484) on port 443, where they blend with web traffic. DoT (RFC 7858) uses port 853, which is easier to block. Block outbound ports 53 and 853 to every resolver except yours. Block known DoH endpoints by domain. Answer Mozilla's canary domain, use-application-dns.net, with NXDOMAIN so Firefox keeps default DoH off, as Mozilla documents.

Connections that never use DNS

Malware that connects to a raw IP address makes no lookup, so there is nothing to block. Firewall egress rules and IP reputation cover that gap.

Domain-level granularity

A resolver sees the domain, never the path. It can block a whole domain but not one page on it. Shared platforms such as cloud document services are all-or-nothing at this layer. A secure web gateway handles page-level decisions.

Data hidden in the lookups

Attackers can encode data in subdomain labels and send it out through ordinary-looking queries. Spotting that needs pattern analysis rather than a blocklist. Our guide to DNS tunnelling detection covers the signals.

Caching and timing

Answers are cached for their time-to-live (TTL). A newly blocked domain stops resolving at the protective resolver straight away. A device may still hold a cached answer until its TTL expires. New threats also appear before feeds list them, which is why domain age matters. See blocking malware before the blocklists catch up.

False positives and failure modes

Every category engine misfiles some domains, so you need a fast allowlist per network. Decide what happens when the resolver is unreachable. A service that fails open hands out unfiltered answers without warning. A service that fails closed keeps policy intact, so it must run across more than one data centre.

How protective DNS relates to Purple Shield

Purple Shield is our protective DNS and content-filtering service. It acts at resolution, before any connection. It catches phishing, malware and command-and-control domains that a firewall only meets after the lookup succeeds. In the terms of this guide, Shield is the service, it holds the filtering policy and it runs the enforcement engine.

Standalone or alongside Purple Access

You can run Shield standalone on any WiFi network or add it to Purple Access. Either way, you set a policy per network and per VLAN. Shield is hardware-agnostic. It works with Cisco Meraki, HPE Aruba, Ruckus, Juniper Mist, Ubiquiti UniFi, Cambium, Extreme and Fortinet, because the change is a DNS setting rather than a new appliance. The Shield how-it-works page walks through setup.

Built to fail closed

Shield fails closed and runs across multiple data centres. Devices never fall back to unfiltered resolution. Shield sits on the Purple platform, which carries 500 million logins a year at a 99.999% uptime SLA.

  • ~241,000threats blocked a weekSt George's Healthcare NHS Trust
  • 80,000+live venuesAcross 90 countries
  • 500 millionlogins a yearPurple platform
  • 99.999%uptime SLAPurple platform

Where the difference shows

Cisco Umbrella and Shield share the same placement. The difference is where policy lives. With Shield, DNS policy sits on the same platform as your Guest WiFi, Staff WiFi and Multi-Tenant WiFi. It maps to networks and VLANs you already manage, across 80,000+ venues in 90 countries.

The concept bites at a specific moment. A guest taps a lookalike login page, or a staff laptop calls a domain registered yesterday. Protective DNS decides in that instant, before anything connects. For the wider picture, read the enterprise WiFi security guide or return to the protective DNS hub. Next, read newly registered domain blocking and why age matters, then stopping phishing at the DNS layer.

Key terms

Protective DNS (PDNS)
A recursive resolver service that checks each query against threat intelligence on malware, phishing, command-and-control and newly registered domains, and refuses to resolve the malicious ones. The term comes from government services run by the UK NCSC and US CISA. You meet it when choosing a DNS security product. It names the security goal, so confirm which threat categories are actually switched on per network.
DNS filtering
The policy practice of allowing or blocking DNS lookups by rule, covering security categories and acceptable-use categories such as adult content or gambling. Protective DNS is DNS filtering with a security mandate. Guest WiFi content filtering is DNS filtering. A content-only configuration leaves security categories off, which is one of the commonest gaps.
DNS firewall
The enforcement mechanism inside resolver software: rules match a query or response and rewrite it to NXDOMAIN, a block page address or a sinkhole. It is commonly built on Response Policy Zones. Vendors use the term for software rules, not an appliance. It can run on a server, a firewall's DNS proxy or a cloud service.
Response Policy Zones (RPZ)
An open format for DNS firewall rules, specified in an IETF Internet-Draft rather than an RFC, and supported in BIND, Unbound and PowerDNS Recursor. If you run your own resolvers, RPZ is how you load threat feeds and policy. It is the shared mechanism behind many commercial services.
Recursive resolver
The DNS server that turns a name into an IP address on a device's behalf, following the resolution process defined in RFC 1034 and RFC 1035. The lookup completes before any TCP or QUIC connection opens. It is the one control point every device on your WiFi touches, which is why a decision here stops traffic before a firewall sees a packet.
DHCP option 6
The DHCP option, defined in RFC 2132, that hands a joining device the addresses of the DNS resolvers it should use. Repointing option 6 is how you deploy protective DNS across guest phones, staff laptops, smart TVs and payment terminals without hardware or agents.
NXDOMAIN
The DNS response stating that a domain does not exist, part of the DNS protocol defined in RFC 1035. Filtering resolvers return it, or REFUSED, to block a lookup. Blocked domains often return NXDOMAIN. It is also the answer you give Mozilla's canary domain to keep Firefox's default DoH off.
Extended DNS Errors
RFC 8914 lets a resolver attach a reason code to a response, including code 15 (Blocked) and code 17 (Filtered). Support for these codes helps your helpdesk tell a policy block from a broken domain when users report sites failing to load.
DNSSEC
DNS Security Extensions, introduced in RFC 4033, add signatures that let a validating resolver prove an answer is authentic. It provides integrity, not policy. DNSSEC will validate a correctly signed phishing domain. You need it alongside protective DNS, and neither replaces the other.
DNS over TLS (DoT)
Encrypted DNS transport specified in RFC 7858, running on port 853 between device and resolver for privacy in transit. Because it uses a dedicated port, DoT to resolvers other than yours is easy to block at the network edge.
DNS over HTTPS (DoH)
Encrypted DNS transport specified in RFC 8484, carried over HTTPS on port 443 where it blends with web traffic. A browser with DoH switched on can bypass your filter entirely, so block known DoH endpoints by domain before relying on DNS policy.
VLAN
A virtual LAN: a logically separate network running on shared hardware, defined by IEEE 802.1Q tagging. A protective resolver maps each query to its VLAN, so guest, staff and payment networks can get three different answers for the same domain.
Sinkhole
A controlled address that a DNS firewall returns in place of the real one, absorbing traffic meant for a malicious domain. Sinkholing is one of the rewrite options a DNS firewall uses, alongside NXDOMAIN and block page addresses.
Fail closed
A failure mode in which an unreachable filtering service does not hand out unfiltered answers, keeping policy intact. A fail-open design does the opposite without warning. Ask every vendor which mode it uses. Fail closed only works if the service runs across more than one data centre.

Worked examples

A venue runs guest WiFi, staff WiFi and a network for payment terminals and IoT devices on one estate. How should protective DNS policy differ across them?

Set three policies, one per VLAN, because the resolver knows which VLAN each query came from. The guest VLAN gets security categories plus adult, gambling and piracy blocks, with a block page that explains why. The staff VLAN gets security categories, newly registered domain blocking and phishing protection, with fewer content blocks, since staff WiFi is where phishing lands. The payment and IoT VLAN gets security categories plus a tight allowlist of only the domains those devices need. That narrows what a compromised device can reach on a network you already segment to reduce PCI DSS scope. The same domain can then receive three different answers.

You run Cisco Umbrella today and want to move to another protective DNS service without leaving devices unprotected.

Treat migration as a staged resolver change. First export your allowlists, blocklists and category settings, then recreate them per network or VLAN in the new service. Repoint DHCP or your local DNS forwarders on one pilot network only, and compare block logs between old and new before moving the rest of the estate. This exposes category differences and false positives while the impact is contained. Remove old agents and forwarders last, so no device keeps resolving through the previous service after cutover. Before signing, ask each vendor to price the identical scope, including support.

Frequently asked questions

Is protective DNS the same as DNS filtering?

No, protective DNS is one use of DNS filtering. DNS filtering is the practice of allowing or blocking lookups by policy, for security or acceptable use. Protective DNS applies that practice with a security mandate, blocking malware, phishing and command-and-control domains. Purple Shield and Cisco Umbrella offer both. Check which categories are switched on in each network's policy before you assume you are protected.

Do I need new hardware to add protective DNS to my WiFi?

No. Protective DNS works by changing the resolver your networks hand out, through DHCP or your local DNS forwarders. Purple Shield is hardware-agnostic. It runs on Cisco Meraki, HPE Aruba, Ruckus, Juniper Mist, Ubiquiti UniFi, Cambium, Extreme and Fortinet networks. You can run it standalone or with Purple Access, with a separate policy for each network and VLAN.

How is protective DNS priced?

Pricing depends on scope, so compare quotes on an identical estate. Variables include venues, networks and VLANs, the categories you need, log retention and whether filtering comes standalone or bundled with access control. Purple quotes Shield standalone or alongside Purple Access. Ask Cisco Umbrella and Purple to price the same scope, including support, before you compare.

How do we migrate from Cisco Umbrella?

Migration is a resolver change, run in stages. Export your allowlists, blocklists and category settings, then recreate them per network or VLAN. Repoint DHCP or forwarders on one pilot network and compare block logs before moving the rest. Remove old agents and forwarders last, so no device keeps resolving through the previous service.

Does protective DNS help with GDPR and PCI DSS?

It supports compliance but does not deliver it alone. DNS logs can count as personal data under UK GDPR when they link to a device or person, so set retention and access controls. On payment networks, a tight allowlist narrows what a compromised terminal can reach within your PCI DSS segmentation. Purple is ISO 27001 certified and GDPR compliant.

What happens if the protective DNS service goes down?

With Purple Shield, devices never fall back to unfiltered lookups, because Shield fails closed. Resilience comes from running across multiple data centres, so another site answers if one is unreachable. Ask every vendor how its service behaves on failure. A fail-open design keeps devices online but removes filtering without warning, which leaves your networks exposed.

Sources

Decide what your network reaches, free for 30 days

Tell us your networks and who uses them, and we map a policy to each. Point one at Shield and see what it stops for 30 days.

Tell us your networks. We map a policy to each one.