How Often Should You Scan External Assets? (By Tier)

How Often Should You Scan External Assets? (By Tier)

Ask “how often should you scan external assets” and someone will usually hand back a single number: weekly, nightly, quarterly. The number is already wrong, because “scan” bundles three separate activities that cost different amounts, carry different risk and answer different questions. Separate them and the schedule more or less writes itself.

“How often” is really three questions, not one

Discovery is finding names, addresses and services you did not know you owned. Observation is looking at what is there: banners, software versions, open interfaces, certificates, DNS and mail configuration. Active testing is proving a weakness actually exists on a live system, with authorisation and a rollback plan.

The cost curve across those three is steep. Certificate transparency logs, passive DNS, registration data and public scan datasets can be queried daily for next to nothing, and they do not touch your systems at all. So there is rarely a defensible reason to discover monthly. Observation sits in the middle: cheap, non-intrusive, but it generates findings that need triage. Active testing is where authorisation, change windows, supplier permission and budget actually bite, and where a careless Tuesday afternoon can take a payment path offline.

Which produces an awkward sequence that a single blanket frequency hides completely. An asset is discovered on Tuesday because a certificate was issued for it. On Wednesday, observation fingerprints it as running a version with a known critical flaw. The flaw is not confirmed until the next authorised test, which may be six weeks away. Three cadences, three different gaps, one risk. If your schedule says “we scan monthly”, nobody can tell which of those gaps is the one you are living with.

Worth being blunt about the overlap with internal tooling. Agent-based vulnerability management and patch reporting already cover your managed estate, often better than any external view can, because they see installed packages rather than inferring them. What they do not see is the agency-hosted microsite on your domain, the CNAME pointing at a SaaS tenant somebody cancelled, the cloud resource in an account that was never enrolled, or the supplier portal branded with your logo. Those are discovery problems, and no agent will ever report them. The difference between inference and proof is covered in more depth in our piece on likely versus definitely.

A tiered scanning cadence for external assets

Tier the estate first. Cadence applied evenly to everything is either too expensive at the top or pointless at the bottom.

Tier 1: revenue and authentication paths

Customer portals, VPN and remote access gateways, single sign-on endpoints, payment flows, anything where a compromise means fraud, data loss or a trading stop. Discovery and observation daily. Active testing monthly, and again on any material change, not instead of it. These are the assets attackers check first because the reward is immediate, and they are also the assets most likely to have an emergency configuration change applied at 4pm on a Friday.

Tier 2: supporting production

Marketing sites, public APIs that do not front money or credentials, mail and DNS infrastructure, status pages, documentation portals. Observation daily where it is cheap, weekly at minimum. Active testing quarterly. Mail and DNS deserve a note: SPF, DKIM and DMARC records fail quietly, and a record loosened for a one-off bulk send in March will still be loose in November unless something checks.

Tier 3: the forgotten estate

Staging and UAT environments, old campaign microsites, third-party hosted subdomains, cloud resources from a project that finished two years ago. Discovery continuous, because the risk here is existence rather than patch level. Nobody is patching a staging box that nobody remembers, so the useful question is not “is it current” but “is it still there, and does it still have a database copy behind it”. The ordinary sediment of a business that has moved quickly lives in this tier, and it is where most genuine surprises come from.

Tier 4: inherited and unverified assets

Anything arriving from an acquisition, a rebrand, an agency handover or a departed supplier. Give these a one-off deep pass: discovery, observation, ownership confirmation, then active testing once somebody has agreed in writing that you are allowed to touch it. After that, fold each asset into one of the three tiers above and stop treating “inherited” as a permanent category. Assets that sit in Tier 4 for a year are assets nobody has owned for a year.

If nobody has tiered anything before, three questions get you most of the way in an afternoon: can it be reached without credentials, what data does it hold or front, and does losing it stop you trading. Two strong yeses put an asset in Tier 1. It will not be perfect. A rough tiering reviewed quarterly beats an elegant one that never gets finished.

Change triggers that should override the calendar

A calendar cadence is the baseline you fall back on when nothing is happening. Something is usually happening.

  • An unfamiliar certificate appears in transparency logs. A certificate issued for a name on your domain that nobody in the team recognises is a same-day investigation, not a note for the next monthly sweep. It is an inference, not proof of anything, but it is the single highest-signal discovery event you get.
  • A cloud resource is decommissioned. The dangling DNS window opens the second the bucket, app service or load balancer is released, and the name becomes claimable by whoever registers it next. Removing the DNS record belongs in the decommission checklist, not the scan schedule. Scanning is the safety net that catches what the process missed, which is why we treat dangling records and subdomain takeover as a process failure first and a detection problem second.
  • A new critical vulnerability lands in software you run externally. Rescan for matching fingerprints within a day or two. Waiting for the quarter means competing with everyone who did not wait. Our external readiness checklist for patch waves covers the sequence in more detail.
  • Business events. An acquisition completes. A new agency starts a campaign. A product launches. Remote access is loosened in a hurry because an executive cannot connect from a hotel. Each of those changes the surface faster than any monthly rhythm will notice.
  • Certificate expiry and weak TLS configuration. The failure date is known in advance, which makes a missed expiry the most avoidable outage on this list. Many issuers now use short-lived certificates with automated renewal, and the CA/Browser Forum has agreed a phased reduction in maximum public certificate lifetimes over the coming years, so the renewal cycle is getting shorter rather than longer. Check daily and alert in stages at 30, 14 and 7 days. Quarterly expired SSL certificate monitoring is a report on outages you have already had.

Compliance frequencies are a floor, not a schedule

PCI DSS requires external vulnerability scanning at least quarterly by an approved scanning vendor, plus after any significant change, and the PCI Security Standards Council publishes the programme requirements for those vendors. Read the scope carefully: the obligation covers the in-scope cardholder data environment. Your estate is larger than that, usually by an order of magnitude, and the quarterly rhythm says nothing about the rest of it.

Cyber Essentials expects vulnerabilities in internet-facing software rated critical or high to be patched or mitigated within 14 days of a fix being released. That figure sets a practical ceiling on the useful gap between external checks. If you promise 14 days in policy but verify externally once a quarter, there is a period of up to three months in which no evidence exists that the promise was kept. Most teams never notice the inconsistency until an insurer or an auditor asks for the proof, and the honest answer is that the proof was never collected.

NCSC guidance on vulnerability management and on preparing for disruptive attacks assumes you already know what is exposed; the NCSC’s vulnerability management guidance is worth reading alongside your own cadence. Discovery frequency so sets your realistic response speed. You cannot patch within 14 days an asset you will not discover for 30.

Write the reasoning down, not just the numbers. One page per tier: the cadence, why that cadence, the exceptions, who approved them and when the exception expires. Auditors rarely argue with a documented judgement. They argue with a cadence nobody can explain.

Scanning more often without drowning the team

Raising frequency without change detection multiplies noise, not coverage. A daily scan that reposts the same 400 findings every morning teaches everybody to close the tab, and a team that has learned to ignore the report is worse off than one that gets a careful quarterly PDF. That is the honest test of whether you are ready to go daily: does the tooling report deltas and route them to a named owner, or does it just repost state?

Four habits make higher frequency survivable:

  • Alert on change, report on state. The daily email contains what is new, what disappeared and what altered. The full inventory stays in a dashboard for when somebody needs it.
  • Suppress with an expiry date. Accepted risk should resurface for review in 90 days, not vanish. Permanent suppression is how a known exception becomes an unknown one.
  • Route by owner at detection. For third-party hosted assets your only available action is a supplier ticket, so the workflow should produce one rather than an internal task nobody can complete. Sorting out who actually owns each third-party subdomain is usually the bottleneck, not the scanning.
  • Split “new asset appeared” from “known asset changed”. Different people, different urgency, different questions. A new name on your domain is an ownership question. A changed banner on a known host is a patching question.

This is also where continuous threat exposure management earns its keep as a practice rather than a label: constant observation to find candidates, prioritisation against what the business actually cares about, then scheduled validation. Our longer argument for continuous monitoring over point-in-time scanning sets out what the snapshot approach structurally cannot see.

Writing your cadence down: a starting plan to adapt

A rhythm you can hand to a new starter beats an intention you hold in your head.

  • Daily: discovery across certificate transparency, DNS and registration data; observation of Tier 1 and Tier 2; certificate expiry and TLS configuration checks; delta alerts only.
  • Weekly: triage of new assets and ownership assignment; review of unresolved deltas; check that decommission tickets closed their DNS records.
  • Monthly: active testing of Tier 1; review of suppressions due to expire; confirmation that Tier 4 items have moved into a real tier.
  • Quarterly: active testing of Tier 2; full inventory review against finance and procurement records for domains, hosting and SaaS; re-tier anything whose business role has changed.

Per asset, record six things: owner, tier, hosting party, last discovered, last observed, last actively tested. Six fields survive staff turnover. A clever taxonomy does not.

Raise frequency when the evidence says to: repeated surprise assets at discovery, frequent cloud churn, a development team shipping new services weekly, an acquisition in progress. Raising it will not help when the backlog is already longer than the team can clear, when findings have no owner, or when the real failure is that nobody removes DNS records at decommission. Frequency does not fix ownership gaps. And no cadence, however tight, turns passive observation into proof; it only shortens the wait before someone confirms it properly.

The Short Version

  • Split the question into discovery, observation and active testing. Discovery is cheap enough to run daily; active testing is the one constrained by authorisation and change windows.
  • Tier 1 revenue and authentication paths: observe daily, test monthly or on change. Tier 2: observe daily to weekly, test quarterly. Tier 3 forgotten estate: continuous discovery, because existence is the risk. Tier 4 inherited: one deep pass, then assign a real tier.
  • Change triggers beat the calendar: unfamiliar certificates, decommissioned cloud resources, new critical vulnerabilities in software you expose, acquisitions and launches.
  • PCI DSS quarterly scanning covers the cardholder data environment only, and a 14-day patching promise cannot be evidenced by a quarterly check.
  • Go daily only if your tooling reports deltas and routes them to an owner. Otherwise you are training people to ignore the report.

Frequently Asked Questions

How often should you scan external assets if you have no inventory yet?

Start with discovery rather than scanning. Run continuous or daily discovery from your known domains and IP ranges for the first few weeks, because the inventory will keep growing and a single pass will miss most of the long tail. Once the curve flattens and new names stop appearing weekly, tier what you have and apply the cadences above.

Is daily scanning of external assets overkill for a small organisation?

Daily discovery is not, because certificate transparency and DNS queries are cheap and need no change window. Daily active testing almost certainly is, and for a small estate monthly or quarterly testing of the assets that matter is a reasonable position to defend. The constraint is usually triage capacity, not scan capacity.

Does more frequent scanning increase the risk of breaking a production system?

Discovery and passive observation carry close to zero operational risk, since they read public records and look at responses rather than exercising inputs. Active testing is the part that can disrupt fragile applications, legacy appliances and rate-limited APIs. Keep active testing on an agreed schedule with a named contact and a rollback plan, and raise the frequency of the passive work instead.

How quickly should you rescan after a new critical vulnerability is published?

Within a day or two for anything you are known to expose, and sooner if exploitation is already being reported. The realistic constraint is whether your inventory lets you answer “do we run this externally” in minutes rather than days. If the answer takes a week to assemble, the gap is in discovery and asset records, not in scan frequency.

Do quarterly PCI DSS scans count as enough external scanning?

They satisfy that particular obligation for the in-scope cardholder data environment, and nothing beyond it. The rest of your internet-facing estate, including the subdomains, staging environments and supplier-hosted services outside PCI scope, is not covered by those scans at all. Treat the quarterly requirement as a floor you have to meet, then build the cadence your actual surface needs on top of it.