ReferenceProtective DNS
DNS filtering logs and reporting: what to record and who reads it
Use this reference to specify the 10 fields a protective DNS resolver should log, decode the response codes behind each block and assign each report to the team that acts on it. Each value traces to RFC 1035, RFC 8914, PCI DSS v4.0.1 Requirement 10, ISO/IEC 27001:2022 or the CCPA/CPRA, so you can set retention and...
Watch the explainer
Every DNS filtering log entry should carry a timestamp, client identifier, queried domain, record type and response code. Add the filtering action and the network or VLAN of origin. PCI DSS v4.0.1 Requirement 10.2.2 sets the audit baseline. IT security reads threats daily, network administrators read faults, and venue operations managers read category trends.
How do you read this DNS filtering logs reference?
This reference covers three things as of June 10, 2026: the fields a protective DNS resolver should log, the response codes behind each block, and who reads the reports built from them.
Start with placement. A protective DNS resolver sees each lookup before the device opens any connection. A firewall sees only the connection that follows. If the lookup was blocked, the firewall sees nothing at all. DNS query logs therefore often hold the earliest record of a phishing click or a malware callback. Purple Shield and Cisco Umbrella both filter at this point, so both work from the same query-level signal.
Each matrix follows three rules:
- Every cell has a source. The Source column names the standard or regulation that defines the value. We cite RFCs, PCI DSS, ISO/IEC 27001 and the CCPA/CPRA, never vendor marketing.
- No guessed values. Where no primary source supports a value, we omit the cell rather than estimate it. Vendor retention limits are left out because they change with each plan and release.
- Dated. Each table carries an as-of date. Check it against the source before you quote a value in a policy or audit response.
Read the main matrix row by row when you specify a resolver or review a log export. Read the second matrix when a DNS filtering report shows a spike in one response code and you need to know what it means.
Flow showing a device lookup reaching the protective resolver, a policy decision to allow or block, a log record and the reports each team reads.
Connections
- Device joins SSID or VLAN to Lookup reaches the resolver
- Lookup reaches the resolver to Policy check
- Policy check to Allowed (allow)
- Policy check to Blocked (block)
- Allowed to Log record written
- Blocked to Log record written
- Log record written to Reports
What should a DNS query log record?
The main matrix lists 10 fields. Five come from the DNS message format in RFC 1035 and its extensions. Four map to the audit attributes in PCI DSS v4.0.1 Requirement 10.2.2. Those attributes are identification, event type, date and time, success or failure, origin, and the affected resource. The last field ties each lookup to a network segment.
| Field | What it records | Read by | Source |
|---|---|---|---|
| Timestamp | Date and time in UTC, from a synchronized clock | IT security manager, auditor | PCI DSS v4.0.1 10.2.2, 10.6.1 |
| Client identifier | Source IP address or device ID that sent the query | IT security manager, DPO | PCI DSS v4.0.1 10.2.2; CCPA/CPRA |
| Query name (QNAME) | The domain name looked up | IT security manager, network administrator | RFC 1035 s4.1.2 |
| Query type (QTYPE) | Record type, such as A, AAAA or HTTPS | Network administrator | RFC 1035 s3.2.2; RFC 3596; RFC 9460 |
| Response code (RCODE) | NOERROR, NXDOMAIN, SERVFAIL or REFUSED | Network administrator | RFC 1035 s4.1.1 |
| Filtering action | Allowed or blocked, with any reason code returned | IT security manager, venue operations manager | PCI DSS v4.0.1 10.2.2; RFC 8914 |
| Category | Event type behind the action, such as malware or phishing | Venue operations manager, IT security manager | PCI DSS v4.0.1 10.2.2 |
| Network and VLAN | SSID or VLAN of origin, and the policy applied | Network administrator, venue operations manager | IEEE 802.11; IEEE 802.1Q |
| Transport | Port 53 DNS, DNS over TLS or DNS over HTTPS | IT security manager | RFC 1035 s4.2; RFC 7858; RFC 8484 |
| Answer data | Records returned, such as the resolved IP address | IT security manager | RFC 1035 s4.1.3 |
The fields are only half the reference. The other half is who reads them, and how often.
What the IT security manager reads
You read blocked threats by category, client and time. PCI DSS v4.0.1 Requirement 10.4.1 requires daily review of security event logs for systems in scope. Look for one client querying several newly blocked domains in a short window. That pattern often points to an infected device. Match the client identifier against Dynamic Host Configuration Protocol (DHCP) lease records to name the device, then isolate it.
What the network administrator reads
You read response codes and query types. A rise in SERVFAIL points to an upstream resolution fault. A rise in REFUSED can mean a DHCP scope is handing out a resolver that does not accept that network. Query types show what devices actually ask for, such as AAAA records for IPv6 or HTTPS records under RFC 9460. The guide to deploying protective DNS per SSID covers the DHCP and resolver mapping behind these fields.
What the venue operations manager reads
You read category trends by network. You want to know which SSIDs see gambling, adult or malware blocks, and whether that matches your acceptable use terms. A virtual LAN (VLAN) is a logical network segment defined by IEEE 802.1Q. Shield applies a policy per network and VLAN, so reports split the same way. A hotel separates guest and staff blocks. An airport separates passenger and operational networks. Whitbread and AGS Airports run Shield.
What the DPO and auditor read
Your data protection officer (DPO) reads retention, access and minimization. Your auditor reads completeness and protection. PCI DSS v4.0.1 Requirement 10.3.1 limits read access to staff with a job-related need. Requirement 10.3.2 protects log files from modification. Requirement 10.3.3 requires prompt backup to a central log server that is difficult to modify.
Which notes change a cell in the DNS query log matrix?
The matrix holds for a single network with one resolver and plain DNS. Real venues add network address translation, encrypted DNS, ad blocking and regulatory scope. Each note below changes how you read one or more cells.
Client identifier behind NAT
Network address translation (NAT) rewrites private device addresses to one public address at the gateway. If the resolver receives lookups from that public address, the client identifier names the site, not the device. To attribute a block to a device, correlate the timestamp with your gateway's NAT and DHCP logs. Per-network policy still applies, because each network is identified separately.
Timestamps across sites
Store every timestamp in UTC. Multi-site estates otherwise produce reports that drift by an hour twice a year. PCI DSS v4.0.1 Requirement 10.6.1 requires synchronized system clocks. ISO/IEC 27001:2022 control A.8.17 requires the same. The Network Time Protocol is defined in RFC 5905.
Encrypted DNS changes the transport cell
DNS over TLS (DoT, RFC 7858) uses port 853. DNS over HTTPS (DoH, RFC 8484) runs inside ordinary HTTPS traffic. A device that sends lookups to its own DoH resolver never appears in your DNS query logs. A client that is active on the network but silent in the resolver log is a signal in itself. The DNS security checklist for venues covers the gateway rules that close this gap.
Ad blocking changes the volume baseline
Ad blocking in Shield reduces data usage and DNS query volume. If you switch it on, query volume reports drop sharply overnight. Reset your baseline after the change, or the drop will read as an outage. The guide to DNS filtering latency covers what the change means for browsing speed.
Guest networks and PCI DSS scope
PCI DSS Requirement 10 applies to systems in, or connected to, the cardholder data environment. If guest WiFi is segmented from payment systems, PCI DSS retention may not apply to guest logs. Staff WiFi that shares a segment with payment terminals brings those logs into scope. Confirm scope with your Qualified Security Assessor (QSA) before you set retention.
CCPA/CPRA minimization changes the client identifier cell
CCPA/CPRA names IP addresses as online identifiers. Compliance requires data minimization and storage limitation. On guest networks, keep full client detail for a short investigation window. Then pseudonymize it or aggregate it for trend reporting.
Fail-closed resolvers and log gaps
Shield fails closed and runs across multiple data centers. If filtering cannot be applied, lookups do not fall through to an unfiltered path. A fault therefore shows up as failed or refused lookups in the log, not as an unfiltered, unlogged window.
How does a blocked lookup appear in the DNS response?
A filter can answer a blocked query in several ways. Each produces a different response code. The second matrix lists the codes you will see in a DNS filtering report and what each standard defines them to mean.
| Code | Name | Defined meaning | Source |
|---|---|---|---|
| RCODE 0 | NOERROR | No error condition; an answer was returned | RFC 1035 s4.1.1 |
| RCODE 2 | SERVFAIL | Server could not process the query because of a server problem | RFC 1035 s4.1.1 |
| RCODE 3 | NXDOMAIN | The queried domain name does not exist | RFC 1035 s4.1.1; RFC 2308 |
| RCODE 5 | REFUSED | Server refuses the operation for policy reasons | RFC 1035 s4.1.1 |
| EDE 15 | Blocked | Domain is on a blocklist set by the resolver operator's own security policy | RFC 8914 |
| EDE 16 | Censored | Domain is on a blocklist required by an entity other than the operator | RFC 8914 |
| EDE 17 | Filtered | Domain is on a blocklist the client asked the resolver to apply | RFC 8914 |
| EDE 18 | Prohibited | Client is not authorized to query this resolver | RFC 8914 |
Why the response code alone undercounts blocks
A filter can return NXDOMAIN, REFUSED, or a NOERROR answer that points to a block page. If your report counts only NXDOMAIN, you miss the other two. Count blocks from the filtering action field in the main matrix instead. Use the response code to diagnose faults, not to measure protection.
What Extended DNS Errors add
Extended DNS Errors (EDE) give a reason alongside the response code. The four filtering codes separate an operator's security block (15) from an external requirement (16). They also separate a client-requested filter (17) from an unauthorized client (18). Not every resolver sends EDE and not every client reads it. Treat EDE as extra detail, never as your only block count.
Which standards and sources sit behind each value?
Every cell in both matrices traces to one of the sources below. Retention and review rules differ by framework, so this section also notes what each one requires.
- RFC 1035 defines the DNS message format, QNAME, QTYPE, RCODE and answer records.
- RFC 2308 covers negative responses, including NXDOMAIN.
- RFC 6891 defines EDNS(0), the extension mechanism that carries EDE.
- RFC 8914 defines Extended DNS Errors, including codes 15 to 18.
- RFC 7858 and RFC 8484 define DNS over TLS and DNS over HTTPS.
- RFC 9499 sets current DNS terminology.
- PCI DSS v4.0.1 Requirement 10 sets the audit attributes (10.2.2) and log protection (10.3). It requires daily review of security event logs (10.4.1). It also sets retention of 12 months, with three months immediately available (10.5.1), and time synchronization (10.6).
- ISO/IEC 27001:2022 Annex A controls 8.15 (logging), 8.16 (monitoring activities) and 8.17 (clock synchronization). The standard sets no fixed retention period; your logging policy defines it.
- CCPA/CPRA Recital 30 (online identifiers), Article 5 (minimization and storage limitation) and Article 32 (security of processing). It sets no fixed retention period, but it requires a justified one.
- NIST SP 800-92, the Guide to Computer Security Log Management, and NIST SP 800-81-2, the Secure Domain Name System Deployment Guide.
For the wider picture, read the protective DNS overview and the enterprise WiFi security guide. Public sector teams should also read NCSC Protective DNS and the UK public sector. To see where Shield sits in your network and how its policies map to each SSID, read how Shield works. Shield runs on the Purple platform, which carries 500 million logins a year at a 99.999% uptime SLA across 80,000+ venues in 90 countries.
How this reference is maintained
We review this reference whenever a cited source is revised. That includes a new RFC, a PCI DSS release or an ISO/IEC 27001 amendment. When a cell changes, we reset the as-of date on that table. When a cell loses its source, we remove it rather than carry it forward. Check the as-of date on each table before you rely on a value.
Key terms
- Protective DNS resolver
- A recursive resolver that applies a filtering policy to each lookup before answering, using the DNS message format in RFC 1035, with terminology set by RFC 9499. It returns an allowed answer or a block response and logs the decision. You meet it when you point each network's DHCP scope at the filtering resolver. It sees the lookup before any connection opens, so its log often holds the earliest record of a phishing click or malware callback.
- RCODE (response code)
- The 4-bit response code in the DNS header defined in RFC 1035 section 4.1.1: 0 NOERROR, 2 SERVFAIL, 3 NXDOMAIN and 5 REFUSED. RFC 2308 covers negative responses such as NXDOMAIN. Network administrators read RCODE trends to diagnose faults. A SERVFAIL rise points upstream and a REFUSED rise can mean a DHCP scope hands out the wrong resolver. RCODE alone undercounts blocks.
- QNAME and QTYPE
- The query name and query type fields of the DNS question section, defined in RFC 1035 sections 4.1.2 and 3.2.2. RFC 3596 adds the AAAA type for IPv6 and RFC 9460 adds the HTTPS record type. QNAME tells IT security which domain a client tried to reach. QTYPE shows network administrators what devices actually request, such as AAAA or HTTPS records.
- Extended DNS Errors (EDE)
- An EDNS(0) option defined in RFC 8914 that carries a reason code alongside the RCODE. Filtering codes are 15 Blocked (operator policy), 16 Censored (external requirement), 17 Filtered (client-requested) and 18 Prohibited (unauthorized client). EDE tells you why a lookup was blocked, not just that it failed. Not every resolver sends it and not every client reads it, so treat it as extra detail, never as your only block count.
- EDNS(0)
- The extension mechanism for DNS defined in RFC 6891. It adds an OPT pseudo-record that carries options such as the Extended DNS Errors defined in RFC 8914. If your resolver or reporting tool does not parse EDNS(0) options, EDE reason codes will be missing from your DNS filtering reports.
- DNS over TLS (DoT) and DNS over HTTPS (DoH)
- Encrypted DNS transports. DoT, defined in RFC 7858, uses port 853. DoH, defined in RFC 8484, carries DNS inside ordinary HTTPS traffic. Plain DNS uses port 53 under RFC 1035 section 4.2. The transport field records which one a lookup used. A device sending DoH to its own resolver never reaches your filter, so gateway rules on port 853 and outside resolvers close the gap.
- VLAN (virtual LAN)
- A logical network segment defined by IEEE 802.1Q, which tags Ethernet frames with a VLAN identifier. SSIDs are defined under IEEE 802.11. Shield applies one policy per network and VLAN, so each log entry carries its segment of origin. That lets a hotel separate guest and staff blocks, or an airport separate passenger and operational networks.
- Network address translation (NAT)
- A gateway function that rewrites private device addresses to one public address. Lookups reaching the resolver from that public address identify the site, not the individual device. Behind NAT, the client identifier cell names the site. To attribute a block to a device, correlate the timestamp with your gateway's NAT and DHCP lease records.
- PCI DSS v4.0.1 Requirement 10
- The PCI Security Standards Council requirement for logging and monitoring. It sets audit attributes (10.2.2), log access and protection (10.3.1 to 10.3.3), daily review (10.4.1), 12 months' retention with three months immediately available (10.5.1) and time synchronization (10.6). It applies to systems in, or connected to, the cardholder data environment. Segmented guest WiFi may sit outside scope, while Staff WiFi sharing a segment with payment terminals brings its logs in. Confirm with your QSA.
- Pseudonymization
- Processing personal data so it can no longer be attributed to a person without separately held information, as defined in CCPA/CPRA regulations. It supports minimization and storage limitation. CCPA/CPRA regulations name IP addresses as online identifiers. On guest networks, keep full client detail for a short investigation window, then pseudonymize or aggregate it for trend reporting.
- Network Time Protocol (NTP)
- The clock synchronization protocol defined in RFC 5905. Synchronized clocks are required by PCI DSS v4.0.1 Requirement 10.6.1 and ISO/IEC 27001:2022 control A.8.17. Multi-site estates without synchronized UTC timestamps produce reports that drift by an hour twice a year, which breaks correlation with firewall, NAT and DHCP logs.
Worked examples
During the daily log review, your IT security manager sees one client identifier query several newly blocked malware and phishing domains within a few minutes. The resolver receives lookups from the gateway's single public address.
One client hitting several newly blocked domains in a short window often points to an infected device. Because NAT rewrites private addresses to one public address, the client identifier names the site, not the device. The team correlated the block timestamps, stored in UTC, with the gateway's NAT and DHCP lease records to name the device, then isolated it. Blocks were counted from the filtering action field rather than NXDOMAIN, because the filter can also return REFUSED or a block page answer. Where logs are in PCI DSS scope, this daily review meets Requirement 10.4.1.
You switch on ad blocking in Shield across your guest WiFi. The next morning, query volume reports show a sharp overnight drop and an operations manager asks whether the resolver is down.
Purple's own data shows ad blocking in Shield cuts up to 62% of DNS queries and up to 44% of data, so a sharp fall is the expected result, not an outage. The team reset the volume baseline from the date of the change so later reports compare like with like. To rule out a real fault, they checked SERVFAIL and REFUSED rates in the response code field. Shield fails closed across multiple data centers, so a genuine fault would appear as failed or refused lookups in the log, not as an unfiltered, unlogged window.
Frequently asked questions
Do I need new hardware to collect DNS filtering logs?
No. Protective DNS works at the resolver, so your access points and switches stay as they are. You point the DHCP scope for each network at the filtering resolver, and the resolver logs every lookup. Shield runs standalone on any WiFi network, including Cisco Meraki, HPE Aruba, Ruckus, Juniper Mist, Ubiquiti UniFi, Cambium, Extreme and Fortinet. It can also be added to Purple Access.
How long should we keep DNS query logs?
Keep them as long as your strictest applicable rule requires, and no longer. Where logs fall inside PCI DSS scope, Requirement 10.5.1 sets 12 months, with three months immediately available. CCPA/CPRA sets no fixed period but requires one you can justify. ISO/IEC 27001:2022 leaves the period to your documented logging policy. Record the period and the reason in your records of processing.
Are DNS query logs personal data under CCPA/CPRA?
Yes, when they link a lookup to an IP address or device identifier. CCPA/CPRA names IP addresses as online identifiers that can identify a person. That brings the minimization, storage limitation and security duties. On guest WiFi, keep full client detail for a short investigation window. Then pseudonymize or aggregate it for trend reporting.
How does a DNS filtering report differ from a firewall report?
A DNS filtering report shows lookups, including blocked ones that never became connections. A firewall report shows the connections that followed. Because protective DNS acts first, it records the phishing domain a guest tried to reach even when no traffic followed. The firewall still matters for traffic that skips DNS, such as hard-coded IP addresses. Correlate both by timestamp and client.
Can guests bypass the filter and disappear from the logs?
Some can, by sending lookups over DNS over HTTPS to a resolver they choose. Those lookups never reach your filter, so the device goes quiet in your DNS query logs. At your gateway, block outbound DNS over TLS on port 853. Also block plain DNS to outside resolvers. Treat a device that is active on the network but silent in the resolver log as an anomaly to investigate.
Can we report separately on guest and staff networks?
Yes. Shield applies a policy per network and VLAN, so each log entry carries its network of origin and reports split the same way. Guest WiFi can run strict category blocking while Staff WiFi follows your workplace policy. Each team then reads only the networks it owns. Whitbread and AGS Airports run Shield, and the per-SSID deployment guide covers the policy mapping step by step.
What happens to our reporting history if we move from another DNS filter?
Your history stays with the previous provider, so export it before you change resolvers. Export the fields in the main matrix as CSV and keep them for the retention period you have set. Note the cutover date in your logging policy. Moving to Shield is a DHCP change per network, so you can migrate one site at a time and compare reports side by side.
