NIST CSF 2.0 for websites and web applications: what a security test can evidence, and where it stops
The NIST Cybersecurity Framework (CSF) 2.0 is a set of cybersecurity outcomes published by the US National Institute of Standards and Technology.
If it has reached your business through a contract, a customer questionnaire or a board paper, this guide explains what the Framework is and how it is organised in NIST's own words, which of its outcomes a website or web application takes part in, and a crosswalk that shows, Subcategory by Subcategory, where a security test can add evidence and where only your own records, monitoring and response plans can.
In this guide
CSF 2.0 evidence crosswalk for websites and applications
Use it on your own systems only. Find each Subcategory your customer, contract or board has asked about, put a name against the suggested owner, collect the artefact, then read the last three columns before relying on a test result for that row.
How to use itHide the instructions
Rows whose evidence boundary begins "Organisational record" are ones the website and application tests described on this page cannot evidence. Four rows have you send requests to your website or application as a check (ID.AM-03, PR.AA-03, PR.AA-05 and PR.DS-02) and carry their limits in the last column. In every other row you read a document, setting, list or page of your own, ask the person who holds a record, or write one. The choice of Subcategories and the mapping are this page's own reading, not a NIST mapping or Informative Reference. Whether the evidence meets a request is decided by whoever made it, or by their auditor or assessor, and NIST does not certify CSF implementations. Subcategory wording is quoted from Appendix A of NIST CSWP 29, with NIST's dashes written as commas. Reprinted courtesy of the National Institute of Standards and Technology, U.S. Department of Commerce.
26 rows
CSF 2.0 Subcategory, in NIST's words
Where your website or application meets it
GV.OC-03 (Govern, Organizational Context): "Legal, regulatory, and contractual requirements regarding cybersecurity, including privacy and civil liberties obligations, are understood and managed"
- Where your website or application meets it
- Contracts, customer questionnaires, insurer forms and privacy duties that mention your website, its data or its testing.
- Suggested owner
- Director or principal, with your adviser for any legal question.
- Artefact to keep
- A register of each requirement: where it came from, what it asks of the website or application, and the date it was last read.
- What a test can contribute
- Nothing towards the outcome itself. Where a contract or questionnaire asks for a test, the report's scope as tested shows which systems, roles and dates that test covered.
- Evidence boundary
- Organisational record. Whether a requirement applies to you, and whether it is managed, is for you and your adviser; a test report cannot show either.
- Check you can run, and its limits
- Read-only: read each contract, questionnaire and policy, and record its exact wording about testing, certification or frameworks. Sends no requests.
GV.RR-02 (Govern, Roles, Responsibilities, and Authorities): "Roles, responsibilities, and authorities related to cybersecurity risk management are established, communicated, understood, and enforced"
- Where your website or application meets it
- Who owns the website and application, who approves changes and releases, who holds administrator access, and who can authorise a security test of each system.
- Suggested owner
- Director or principal.
- Artefact to keep
- This crosswalk's owner column, completed and dated, with the name of the person who can sign a test authorisation for each system.
- What a test can contribute
- Nothing towards the outcome. A test's signed authorisation records who approved testing of each target, which is one entry in this record.
- Evidence boundary
- Organisational record. A test cannot show that roles are understood or enforced.
- Check you can run, and its limits
- Read-only: fill in the owner column of this crosswalk and have the director sign and date it. Sends no requests.
GV.SC-04 (Govern, Cybersecurity Supply Chain Risk Management): "Suppliers are known and prioritized by criticality"
- Where your website or application meets it
- Your hosting provider, domain registrar, content management system, plugins, form and booking services, payment provider, email service, and the developer or agency that maintains the site.
- Suggested owner
- Operations or practice manager, with the developer for technical services.
- Artefact to keep
- A supplier list naming what each supplier runs for the website or application and how critical it is to you.
- What a test can contribute
- A test's scope and exclusions list names the third-party services left out of testing, such as a payment page hosted by your payment provider. Compare that list with yours to find suppliers you had not recorded.
- Evidence boundary
- A test does not assess suppliers or rank their criticality. Services run by other businesses are theirs to test.
- Check you can run, and its limits
- Read-only: list every service in your website's plugin or integration list, your invoices and your DNS records, without opening any settings screen that shows a key, password or connection detail, and compare the list with the last test's scope and exclusions. Sends no requests.
GV.SC-07 (Govern, Cybersecurity Supply Chain Risk Management): "The risks posed by a supplier, their products and services, and other third parties are understood, recorded, prioritized, assessed, responded to, and monitored over the course of the relationship"
- Where your website or application meets it
- Vendors whose platforms hold your customers' data or run part of your site, such as a booking engine, a customer management system, a payment provider or a hosted store.
- Suggested owner
- Operations or practice manager.
- Artefact to keep
- Each critical vendor's security documentation, its terms on customer testing, and your dated notes of what you asked it and what it answered.
- What a test can contribute
- If the vendor's terms permit it, or the vendor says yes, a test can take in your half of the integration: the data your application passes to the vendor and the replies it accepts. The vendor's platform is the vendor's to test.
- Evidence boundary
- Testing your application does not assess the supplier. Read the vendor's terms and ask the vendor before including anything it runs.
- Check you can run, and its limits
- Read-only: read each critical vendor's published security and testing terms, record what they say about customer testing, and ask the vendor where they are silent. Sends no requests.
ID.AM-02 (Identify, Asset Management): "Inventories of software, services, and systems managed by the organization are maintained"
- Where your website or application meets it
- Every domain and subdomain, staging copy, API, admin area, plugin, theme and library that makes up the website or application.
- Suggested owner
- Developer or agency, reviewed by the operations manager.
- Artefact to keep
- A dated inventory of hostnames, environments, APIs and installed components, with the person responsible for each.
- What a test can contribute
- An external check lists the hostnames it found answering on your domain, where each points, and the software versions visible from outside; a WordPress audit adds the installed plugin and theme versions. Each is a dated snapshot to compare with your inventory.
- Evidence boundary
- The test lists what it found in its scope on its dates. It does not maintain your inventory, and anything outside its scope is missing from it.
- Check you can run, and its limits
- Read-only: compare your inventory with your DNS records, your hosting account's list of sites and your plugin or package list, and add anything missing. Sends no requests.
ID.AM-03 (Identify, Asset Management): "Representations of the organization's authorized network communication and internal and external network data flows are maintained"
- Where your website or application meets it
- Where a form submission, sign-up or upload goes after it leaves the browser: inboxes, customer management tools, analytics and advertising tags, and any partner it is passed to.
- Suggested owner
- Developer, with whoever handles privacy in your business.
- Artefact to keep
- A data-flow note for each form and upload: each destination, who can read it, and how long it is kept.
- What a test can contribute
- Little towards the outcome. A test keeps the request and response behind each finding as redacted evidence, and records each negative result with the request, account and build it covered. The data-flow note itself is written and kept by the owners this row names.
- Evidence boundary
- Flows that happen on the server, between your own systems or inside a vendor, are not visible from the browser and need the developer's account of them.
- Check you can run, and its limits
- Sends requests. The check: submit each public form once with invented details and an email address you control, then list every inbox, tool and service that received the submission and who can read each. Requests in this row: one submission for each form. Limits on this row: run it only on a website or application your business runs, on a staging copy where one exists. Where it signs in, use only test accounts you created, holding invented records, never a real customer's or staff member's details. No load: send only the requests listed, one at a time. Any email or text it triggers goes to an address or number you control, and a sign-in or reset link is handled like a password: use it only on the test account and record that it arrived, not the link itself. Where a vendor runs any part of the system, its terms decide how far you go: read them first, ask the vendor about anything they leave out, and test no further than the terms or the vendor allow. Should a response show real data belonging to someone else, stop at once, write down the address and the time, go no further, and report it to the person in your business responsible for privacy.
ID.RA-01 (Identify, Risk Assessment): "Vulnerabilities in assets are identified, validated, and recorded"
- Where your website or application meets it
- Weaknesses in the website or application itself: access control, authentication, sessions, input handling, configuration and outdated components.
- Suggested owner
- Technical lead or developer.
- Artefact to keep
- Dated test and assessment reports, a findings list with severity, and a record of any weakness your own team found or someone outside reported.
- What a test can contribute
- A test contributes directly here. A report can record each finding with the evidence that reproduces it, which speaks to "validated", and a rating. NIST's online Implementation Examples for this Subcategory include "Review, analyze, or test organization-developed software to identify design, coding, and default configuration vulnerabilities".
- Evidence boundary
- Covers the systems, roles, build and window tested. Weaknesses in other assets, or introduced after the test, are outside it, and a check that found nothing says nothing about checks that were not run.
- Check you can run, and its limits
- Read-only: read the last report's scope as tested and list every system in your inventory that it did not cover. Sends no requests.
ID.RA-06 (Identify, Risk Assessment): "Risk responses are chosen, prioritized, planned, tracked, and communicated"
- Where your website or application meets it
- What happens to each finding: fixed now, fixed later, accepted, or handled another way.
- Suggested owner
- Director or principal decides; technical lead tracks.
- Artefact to keep
- A findings register with the decision on each finding, who made it, the due date and the status after the retest.
- What a test can contribute
- A report's remediation plan puts the fixes in priority order, and a retest records each finding as fixed, partially fixed, still reproducible or not retested.
- Evidence boundary
- Choosing a response, including accepting a risk, is the business's decision. A finding marked not retested is not evidence that it was fixed.
- Check you can run, and its limits
- Read-only: for each open finding, record the decision, the owner and a date, and update the register after the retest. Sends no requests.
ID.RA-08 (Identify, Risk Assessment): "Processes for receiving, analyzing, and responding to vulnerability disclosures are established"
- Where your website or application meets it
- How someone outside the business who finds a weakness in your site can report it, and what happens to the report.
- Suggested owner
- Technical lead.
- Artefact to keep
- A published way to report a vulnerability, an internal procedure for handling reports, and a log of reports received and answered.
- What a test can contribute
- Nothing towards the outcome. A test report is not a disclosure, and a test does not exercise your process. The vulnerability disclosure policy template in this library covers what a published policy sets out.
- Evidence boundary
- Organisational record. Whether the process works is shown by your log of reports and responses.
- Check you can run, and its limits
- Read-only: open your own public pages as any visitor would, submit nothing, and record whether any of them says how to report a security problem and to whom.
ID.IM-02 (Identify, Improvement): "Improvements are identified from security tests and exercises, including those done in coordination with suppliers and relevant third parties"
- Where your website or application meets it
- The changes you decide to make to the website, the way it is built and released, or its hosting after a test.
- Suggested owner
- Technical lead, reporting to the director.
- Artefact to keep
- A dated note after each test and retest listing the improvements chosen beyond the individual fixes, with owners and dates.
- What a test can contribute
- The test is the input this Subcategory names. NIST's online Implementation Examples for it include "Perform penetration testing to identify opportunities to improve the security posture of selected high-risk systems as approved by leadership".
- Evidence boundary
- The outcome is that improvements are identified from the test, and that record is yours. A report lists findings and fixes, not the changes you choose to make to how you work.
- Check you can run, and its limits
- Read-only: after each test and retest, write down what you will change in how the site is built, released or hosted, with an owner and a date. Sends no requests.
ID.IM-04 (Identify, Improvement): "Incident response plans and other cybersecurity plans that affect operations are established, communicated, maintained, and improved"
- Where your website or application meets it
- What the business does if the website is defaced or taken over, starts serving malware, or exposes customer data.
- Suggested owner
- Director or principal.
- Artefact to keep
- An incident response plan naming who decides, who to call (host, developer, incident response specialist, insurer and adviser) and where the contact list is kept.
- What a test can contribute
- Nothing. A security test of the website does not exercise an incident response plan.
- Evidence boundary
- Organisational record. Exercising the plan, such as a tabletop walk-through, is a separate activity from a security test.
- Check you can run, and its limits
- Read-only: check that the plan names the website and application, the host and developer contacts, and the date it was last reviewed. Sends no requests.
PR.AA-03 (Protect, Identity Management, Authentication, and Access Control): "Users, services, and hardware are authenticated"
- Where your website or application meets it
- Sign-in, sign-up, password reset, second factors and sessions for customers, staff and administrators, and the keys or tokens services use to call your API.
- Suggested owner
- Technical lead or developer.
- Artefact to keep
- The authentication settings for each role, which roles require a second factor, and how API keys are issued and revoked.
- What a test can contribute
- A signed-in test checks sign-in, password reset and sessions for each role in scope, and records what it found and which checks found nothing.
- Evidence boundary
- Covers the application and roles tested. Staff email, cloud consoles and other systems with their own sign-in are outside it.
- Check you can run, and its limits
- Sends requests. The check: sign in once as a test account in each role and record whether a second factor is required, offered or absent, without opening the screen that sets one up; then request one password reset for one test account and record where the reset message arrived. Requests in this row: one sign-in for each role and one reset. Limits on this row: run it only on a website or application your business runs, on a staging copy where one exists. Where it signs in, use only test accounts you created, holding invented records, never a real customer's or staff member's details. No load: send only the requests listed, one at a time. Any email or text it triggers goes to an address or number you control, and a sign-in or reset link is handled like a password: use it only on the test account and record that it arrived, not the link itself. Where a vendor runs any part of the system, its terms decide how far you go: read them first, ask the vendor about anything they leave out, and test no further than the terms or the vendor allow. Should a response show real data belonging to someone else, stop at once, write down the address and the time, go no further, and report it to the person in your business responsible for privacy.
PR.AA-05 (Protect, Identity Management, Authentication, and Access Control): "Access permissions, entitlements, and authorizations are defined in a policy, managed, enforced, and reviewed, and incorporate the principles of least privilege and separation of duties"
- Where your website or application meets it
- The data and actions open to each role, whether one customer's records are reachable by another customer, and who holds administrator access.
- Suggested owner
- Director owns the policy; developer implements it.
- Artefact to keep
- A written permission matrix by role, a dated review of who holds each role, and the record of accounts removed when people leave.
- What a test can contribute
- The enforced part, for the application tested: a signed-in test with two accounts in each role checks access in both directions between users, and calls the API directly as well as through the screens.
- Evidence boundary
- A test cannot show that permissions are defined in a policy or reviewed, or anything about systems outside its scope.
- Check you can run, and its limits
- Sends requests. The check: create two test accounts in the same role, each owning one invented record. Signed in as the first, request the second account's record once by changing its id in the address, and record whether it was refused. Requests in this row: creating the two test accounts and one invented record for each, two sign-ins, and one request for the other record. Limits on this row: run it only on a website or application your business runs, on a staging copy where one exists. Where it signs in, use only test accounts you created, holding invented records, never a real customer's or staff member's details. No load: send only the requests listed, one at a time. Any email or text it triggers goes to an address or number you control, and a sign-in or reset link is handled like a password: use it only on the test account and record that it arrived, not the link itself. Where a vendor runs any part of the system, its terms decide how far you go: read them first, ask the vendor about anything they leave out, and test no further than the terms or the vendor allow. Should a response show real data belonging to someone else, stop at once, write down the address and the time, go no further, and report it to the person in your business responsible for privacy.
PR.DS-01 (Protect, Data Security): "The confidentiality, integrity, and availability of data-at-rest are protected"
- Where your website or application meets it
- The database, file storage, uploads, exports and backups behind the website or application.
- Suggested owner
- Developer, with your hosting or database provider for the controls it runs.
- Artefact to keep
- Your hosting and database provider's documentation on encryption at rest, and your own record of where each copy of the data is stored.
- What a test can contribute
- Indirect. A test can find stored data exposed through the application, such as a backup file or export the web server will serve, or records that a role can reach when it should not.
- Evidence boundary
- A web test cannot see how stored data is encrypted. That evidence comes from your provider's documentation and settings.
- Check you can run, and its limits
- Read-only: read your hosting and database provider's documentation and record what it says about encryption at rest for your plan. Do not open screens that show keys or connection details. Sends no requests.
PR.DS-02 (Protect, Data Security): "The confidentiality, integrity, and availability of data-in-transit are protected"
- Where your website or application meets it
- HTTPS on every public hostname, redirects from HTTP to HTTPS, certificate validity and the Strict-Transport-Security header.
- Suggested owner
- Developer or hosting provider.
- Artefact to keep
- The list of public hostnames, with the certificate and HTTPS settings for each.
- What a test can contribute
- An external check records the certificate and transport security results for each public hostname it covers, with the date.
- Evidence boundary
- Seen from outside only. Traffic between your server and its database or other services is not visible to a web test.
- Check you can run, and its limits
- Sends requests. The check: for each public hostname on your own inventory, open it once in a browser with http:// in front, and record whether it moves to https:// and whether the browser reports a certificate problem. Requests in this row: one for each hostname. Limits on this row: run it only on a website or application your business runs, on a staging copy where one exists. Where it signs in, use only test accounts you created, holding invented records, never a real customer's or staff member's details. No load: send only the requests listed, one at a time. Any email or text it triggers goes to an address or number you control, and a sign-in or reset link is handled like a password: use it only on the test account and record that it arrived, not the link itself. Where a vendor runs any part of the system, its terms decide how far you go: read them first, ask the vendor about anything they leave out, and test no further than the terms or the vendor allow. Should a response show real data belonging to someone else, stop at once, write down the address and the time, go no further, and report it to the person in your business responsible for privacy.
PR.DS-11 (Protect, Data Security): "Backups of data are created, protected, maintained, and tested"
- Where your website or application meets it
- Backups of the website files, database and uploads, wherever your host or developer keeps them.
- Suggested owner
- Hosting provider or developer, checked by the operations manager.
- Artefact to keep
- The backup schedule, where copies are kept, who can reach them, and the date and result of the last test restore.
- What a test can contribute
- Nothing towards the outcome. An external check may find a backup file the web server will serve to anyone, which is a finding about exposure, not evidence that backups work.
- Evidence boundary
- Organisational record. Whether backups exist and restore is shown only by your restore records.
- Check you can run, and its limits
- Read-only: ask your host or developer for the date and result of the last test restore, and file their answer. Sends no requests.
PR.PS-01 (Protect, Platform Security): "Configuration management practices are established and applied"
- Where your website or application meets it
- Security headers, cookie settings, error pages, debug output, exposed administrator paths and staging copies.
- Suggested owner
- Developer.
- Artefact to keep
- A record of the intended configuration for each environment, and the change history that shows how it is kept.
- What a test can contribute
- A test records the configuration it observed on each hostname and build, such as headers, cookie attributes and error handling, and where it differed from the report's recommendation.
- Evidence boundary
- A snapshot of the configuration during the test. It does not show that a practice exists, or that the next release keeps the same settings.
- Check you can run, and its limits
- Read-only: compare the configuration findings in the last report with your change history and note any setting that has changed since. Sends no requests.
PR.PS-02 (Protect, Platform Security): "Software is maintained, replaced, and removed commensurate with risk"
- Where your website or application meets it
- Updates to the content management system, plugins, themes, frameworks and libraries, and removal of unused components.
- Suggested owner
- Developer or agency.
- Artefact to keep
- An update log with dates, and a list of components removed or replaced.
- What a test can contribute
- A test records the versions it could see on the date tested, and a WordPress audit adds an update-priority list for the installed plugins.
- Evidence boundary
- One date's versions. Whether software is kept up to date between tests is shown by the update log.
- Check you can run, and its limits
- Read-only: in your own admin area, read the list of installed plugins or packages and their update status, and record it with the date. Changes nothing.
PR.PS-04 (Protect, Platform Security): "Log records are generated and made available for continuous monitoring"
- Where your website or application meets it
- Application logs of sign-ins, failed sign-ins, refused requests and administrator actions, and the web server and hosting logs.
- Suggested owner
- Developer, with whoever monitors alerts.
- Artefact to keep
- A list of what each system logs, where the logs go, how long they are kept and who can read them.
- What a test can contribute
- Indirect. The test window lets you look afterwards for the tester's refused requests in your own logs.
- Evidence boundary
- A test does not show that logs are kept, protected or watched; those are facts about how you operate.
- Check you can run, and its limits
- Read-only: after an agreed test window, ask whoever runs your logs to search for the test accounts during that window and report how many entries they found and which fields each held, without opening or copying entries that hold passwords, tokens or keys. Sends no requests.
PR.PS-06 (Protect, Platform Security): "Secure software development practices are integrated, and their performance is monitored throughout the software development life cycle"
- Where your website or application meets it
- How the application is built: code review, dependency updates, handling of secrets, and testing before release.
- Suggested owner
- Technical lead.
- Artefact to keep
- Your development process in writing, and records that it is followed, such as pull request reviews and dependency alerts.
- What a test can contribute
- A secure code review reads the source of agreed areas for flaws a running test may not reach. A penetration test of the running application shows the results of the process for the build tested.
- Evidence boundary
- Neither shows that practices are integrated across the life cycle or that their performance is monitored; that comes from your development records.
- Check you can run, and its limits
- Read-only: pick one recent release and find its review and test records. Sends no requests.
PR.IR-04 (Protect, Technology Infrastructure Resilience): "Adequate resource capacity to ensure availability is maintained"
- Where your website or application meets it
- Whether the site stays up under heavy traffic or an attempt to overload it.
- Suggested owner
- Hosting provider and developer.
- Artefact to keep
- Your hosting plan's capacity, the provider's availability terms, and any load test report.
- What a test can contribute
- Nothing from a Clever Ops test, because load and denial-of-service testing are outside every Clever Ops service.
- Evidence boundary
- Evidence here comes from your hosting plan and, if you commission one, a load test by a specialist, run with your host's written approval.
- Check you can run, and its limits
- Read-only: read your hosting plan and the provider's availability terms, and record what they say. Do not run a load test yourself. Sends no requests.
DE.CM-09 (Detect, Continuous Monitoring): "Computing hardware and software, runtime environments, and their data are monitored to find potentially adverse events"
- Where your website or application meets it
- Whether anyone, or any service, watches the application's sign-ins, errors and administrator actions and raises an alert.
- Suggested owner
- Whoever monitors the site: your developer, your host or a monitoring provider.
- Artefact to keep
- What is monitored, who receives alerts, and the record of alerts raised and handled.
- What a test can contribute
- An agreed test window lets you see whether your monitoring raised an alert during it. NIST's online Implementation Examples for this Subcategory include "Monitor authentication attempts to identify attacks against credentials and unauthorized credential reuse".
- Evidence boundary
- Monitoring is continuous and a test lasts its window, so a test is not monitoring. Clever Ops does not provide round-the-clock monitoring.
- Check you can run, and its limits
- Read-only: after the test window, ask whoever monitors the site whether any alert fired, and file the answer with the dates. Sends no requests.
DE.AE-08 (Detect, Adverse Event Analysis): "Incidents are declared when adverse events meet the defined incident criteria"
- Where your website or application meets it
- Deciding when something odd on the website, such as a defacement, an unexpected administrator account or a run of failed sign-ins, becomes an incident.
- Suggested owner
- Director or principal.
- Artefact to keep
- Written incident criteria and the name of the person who declares an incident.
- What a test can contribute
- Nothing. A finding in a test report describes a weakness the test found; deciding what counts as an incident is yours.
- Evidence boundary
- Organisational record.
- Check you can run, and its limits
- Read-only: check that your incident response plan states the criteria and who applies them. Sends no requests.
RS.MA-01 (Respond, Incident Management): "The incident response plan is executed in coordination with relevant third parties once an incident is declared"
- Where your website or application meets it
- Calling the host, developer, incident response specialist and others the plan names when the site is compromised.
- Suggested owner
- Director or principal.
- Artefact to keep
- The plan, its contact list and a record of each incident handled.
- What a test can contribute
- Nothing. Clever Ops does not provide incident response; a site that is compromised now needs an incident response specialist first.
- Evidence boundary
- Organisational record. Response evidence comes from incident records.
- Check you can run, and its limits
- Read-only: check that every contact in the plan is current. Sends no requests.
RS.AN-03 (Respond, Incident Analysis): "Analysis is performed to establish what has taken place during an incident and the root cause of the incident"
- Where your website or application meets it
- Working out how a compromise happened: which weakness, account or supplier was involved.
- Suggested owner
- Incident response specialist, engaged by the director.
- Artefact to keep
- The specialist's findings and the evidence they preserved.
- What a test can contribute
- Nothing. Forensics is not part of any Clever Ops service. When a specialist has finished the cleanup, an audit can look at the web application before customers use it again.
- Evidence boundary
- A later test can find weaknesses that remain; it does not establish what happened during the incident.
- Check you can run, and its limits
- Read-only: confirm the plan names an incident response specialist and how to reach them. Sends no requests.
RC.RP-05 (Recover, Incident Recovery Plan Execution): "The integrity of restored assets is verified, systems and services are restored, and normal operating status is confirmed"
- Where your website or application meets it
- Bringing the website or application back after an incident, and deciding it is fit to take customers again.
- Suggested owner
- Director or principal, on the incident response specialist's advice.
- Artefact to keep
- Recovery records: what was restored, from which backup, and who confirmed normal operation.
- What a test can contribute
- After a specialist has cleaned up, a post-cleanup audit can check what the web application exposes before you rely on it again, for its scope and date.
- Evidence boundary
- Confirming normal operating status is the business's decision, and checking restored systems beyond the web application is outside a web test.
- Check you can run, and its limits
- Read-only: record which backup was used and who confirmed the restoration. Sends no requests.
What the NIST CSF is, in NIST's own words
The NIST Cybersecurity Framework (CSF) 2.0 is published by the National Institute of Standards and Technology as NIST Cybersecurity White Paper 29 (NIST CSWP 29), dated 26 February 2024.
Its abstract says the CSF 2.0 "provides guidance to industry, government agencies, and other organizations to manage cybersecurity risks", and that it offers "a taxonomy of high-level cybersecurity outcomes" that any organisation, whatever its size, sector or maturity, can use to better understand, assess, prioritise and communicate its cybersecurity efforts.
The same abstract says the CSF does not prescribe how those outcomes should be achieved. Instead it links to online resources with further guidance on practices and controls that could be used to achieve them.
NIST is a United States government agency. Its About page says it was founded in 1901 and is part of the U.S. Department of Commerce, and its CSF FAQs say NIST is not a regulatory agency. CSWP 29 describes the CSF as a foundational resource that "may be adopted voluntarily and through governmental policies and mandates".
Asked whether an organisation is required to use the CSF, NIST's FAQ says that Executive Order 13800 made the Framework mandatory for U.S. Federal Government agencies, and that some companies require the CSF for their customers or within their supply chain.
That last route is one way a United States framework can reach an Australian business: a customer, or a business further up a supply chain, names it in a contract or a questionnaire. This page makes no statement that any Australian law or regulator requires the CSF. If someone tells you one does, ask them for the provision, and take it to your adviser.
CSWP 29 says the CSF is designed to be used by organisations of all sizes and sectors, including industry, government, academia and nonprofit organisations, regardless of the maturity of their cybersecurity programs, and that its outcomes are sector-, country- and technology-neutral. Before version 2.0 the Framework had a different title, the "Framework for Improving Critical Infrastructure Cybersecurity", and the publication says that title is not used for CSF 2.0.
How CSF 2.0 is organised: Functions, Categories and Subcategories
CSWP 29 lists the components of CSF 2.0.
The CSF Core is the taxonomy of outcomes. Organizational Profiles describe an organisation's current or target cybersecurity posture, or both, in terms of the Core's outcomes. Tiers can be applied to those Profiles to characterise the rigour of an organisation's cybersecurity risk governance and management practices.
The Core arranges outcomes by Function, then Category, then Subcategory. The publication's glossary says there are six CSF Functions, and Appendix A gives each one a two-letter identifier and a one-line outcome:
- Govern (GV): "The organization's cybersecurity risk management strategy, expectations, and policy are established, communicated, and monitored".
- Identify (ID): "The organization's current cybersecurity risks are understood".
- Protect (PR): "Safeguards to manage the organization's cybersecurity risks are used".
- Detect (DE): "Possible cybersecurity attacks and compromises are found and analyzed".
- Respond (RS): "Actions regarding a detected cybersecurity incident are taken".
- Recover (RC): "Assets and operations affected by a cybersecurity incident are restored".
Reading a Subcategory identifier
Each Category identifier joins its Function's letters to two more, and each Subcategory adds a number. PR.AA-05 is a Subcategory of PR.AA, Identity Management, Authentication, and Access Control, in the Protect Function. Appendix A says the numbering of the Subcategories is intentionally not sequential, and that gaps in numbering indicate CSF 1.1 Subcategories that were relocated in CSF 2.0. So PR.DS-10 following PR.DS-02 is not a typing error.
CSWP 29 also says the Functions should be addressed concurrently, and places Govern at the centre of its Functions wheel because it informs how an organisation will implement the other five.
Outcomes, not a checklist
The publication says the Core's outcomes "are not a checklist of actions to perform", and that the actions taken to achieve an outcome, and the person responsible for them, will vary by organisation and use case.
It adds that the order and size of the Functions, Categories and Subcategories does not imply the sequence or importance of achieving them. NIST's FAQ is just as direct: asked whether the CSF provides a recommended checklist of what all organisations should do, it answers no: the CSF provides a series of outcomes to prioritise and address cybersecurity risks but does not specify actions for meeting those outcomes.
So the CSF itself gives you no checklist to work through. The crosswalk on this page is a checklist of evidence for the parts of the CSF your website or application takes part in, and each row says where that evidence stops.
Profiles and Tiers: where a test result fits and where it does not
CSWP 29 says a Current Profile specifies the Core outcomes that an organisation is achieving or attempting to achieve, and characterises how or to what extent each outcome is being achieved.
A Target Profile specifies the desired outcomes that an organisation has selected and prioritised for achieving its cybersecurity risk management objectives. The publication illustrates one way an organisation could create and use a Profile: scope the Profile, gather the information needed to prepare it, create it, analyse the gaps between the Current and Target Profiles and create an action plan, then implement the plan and update the Profile.
Scope matters here. CSWP 29 says an organisation can have as many Organizational Profiles as desired, each with a different scope, and its examples include a Profile scoped to an organisation's financial systems. Read that way, a Profile can be scoped to the website and application your customers use.
The information gathered at the second step may include, among other things, organisational policies, the cybersecurity requirements and standards the organisation follows, and its practices and tools. A dated test report for the system in scope can be one piece of that information, alongside the records the crosswalk lists.
Profiles also travel between businesses. The publication says a Current Profile can be used to document and communicate an organisation's cybersecurity capabilities and known opportunities for improvement with external stakeholders, such as business partners or prospective customers, and that a Target Profile can help express an organisation's cybersecurity risk management requirements and expectations to suppliers, partners and other third parties as a target for those parties to achieve.
If a customer has sent you a Target Profile as its expectations of suppliers, the Subcategories in it are the outcomes it is asking you about, and the crosswalk shows which of them your website or application evidence can speak to.
Tiers are a different matter. CSWP 29 says Tiers characterise the rigour of an organisation's cybersecurity risk governance and management practices, names them Partial (Tier 1), Risk Informed (Tier 2), Repeatable (Tier 3) and Adaptive (Tier 4), and gives a notional illustration of them in Appendix B.
An organisation can choose to use them to inform its Profiles. A test of one website or application says nothing about which Tier describes a business, because Tiers characterise the organisation's governance and management practices rather than one system, and this page does not rate maturity.
Where a website or web application meets CSF 2.0
CSF 2.0 has no Category for websites or web applications.
CSWP 29 says its Functions, Categories and Subcategories apply to all ICT used by an organisation, including information technology, the Internet of Things and operational technology, and to all types of technology environments, including cloud, mobile and artificial intelligence systems. A website and the application behind it sit inside the Core in the same way as any other system, so the useful question is which outcomes they take part in.
Read against Appendix A, these are the places a customer-facing website or application shows up. The choice of Subcategories is this page's reading, not NIST's.
- Govern: who owns the site, which contracts and questionnaires mention it, and which suppliers run parts of it (GV.OC-03, GV.RR-02, GV.SC-04 and GV.SC-07).
- Identify: the site, its subdomains and its APIs belong in the inventories of ID.AM-02, form data belongs in the data flows of ID.AM-03, and the weaknesses a test reports belong in ID.RA-01, which reads "Vulnerabilities in assets are identified, validated, and recorded".
- Protect: sign-in and sessions (PR.AA-03), what each role can reach (PR.AA-05), data at rest and in transit (PR.DS-01 and PR.DS-02), configuration, updates, logs and how the software is built (PR.PS-01, PR.PS-02, PR.PS-04 and PR.PS-06), and capacity under heavy traffic (PR.IR-04).
- Detect: whether anyone would notice an attack on the login or the API (DE.CM-09), and who decides that something is an incident (DE.AE-08).
- Respond and Recover: what happens when the site is compromised, and how it is brought back (RS.MA-01, RS.AN-03 and RC.RP-05).
What NIST's Implementation Examples say about testing
CSWP 29 does not mention penetration testing. It appears in NIST's online Implementation Examples, which Appendix A says are available on the CSF 2.0 website and through the CSF 2.0 Reference Tool, and which NIST offers as a workbook from its Informative References page.
For ID.RA-01 they include "Review, analyze, or test organization-developed software to identify design, coding, and default configuration vulnerabilities". For ID.IM-02, "Improvements are identified from security tests and exercises, including those done in coordination with suppliers and relevant third parties", they include "Perform penetration testing to identify opportunities to improve the security posture of selected high-risk systems as approved by leadership".
CSWP 29 describes Implementation Examples as notional examples of concise, action-oriented steps, and says they are not a comprehensive list of all actions an organisation could take to achieve an outcome, nor a baseline of required actions to address cybersecurity risks.
The publication also says it does not prescribe outcomes nor how they may be achieved. On that reading, a penetration test is one way to work towards some outcomes, not a step the CSF sets.
A mapping, not an Informative Reference
CSWP 29 calls the mappings that link the Core to other standards, guidelines, regulations and other content Informative References, and says NIST or another organisation may produce them. NIST's Informative References page says NIST does not conduct correctness testing on mappings submitted by others, and that listing a non-NIST mapping in its catalogue does not imply NIST endorsement.
The crosswalk on this page is not an Informative Reference of either kind. It is this page's own reading of where a website or application test can add evidence, not a NIST mapping.
What a website or application test can contribute
A security test report speaks for one system, inside an agreed scope, on the dates it was tested.
Against CSF 2.0 it is an input to a handful of Subcategories and proof of none of them, because each Subcategory is an outcome for the organisation and a report describes one system on its dates. These are the parts of a report that carry the weight.
- Findings with evidence. Every finding in a Clever Ops report comes with a description, its evidence with secrets redacted, what it means for the business, and a fix. That is a record of vulnerabilities identified and validated for the scope tested, which is what ID.RA-01 describes, for that system and those dates only.
- The remediation plan and the retest. The plan ranks the fixes by priority, and the retest then labels each finding fixed, partially fixed, still reproducible or not retested. That feeds ID.RA-06, where the business chooses and tracks its responses, and ID.IM-02, where the business records the improvements it took from the test.
- Negative results. A check that found nothing still goes in the report, tied to the build and accounts it covered. For PR.AA-03 and PR.AA-05, those results show that the sign-in, session and access checks that were run behaved as expected on that build. They do not show that permissions are defined in a policy or reviewed.
- External check results. Certificates and transport security, security headers, cookie settings, visible software versions and exposed files, each recorded for the hostname and date checked. These speak to PR.DS-02, PR.PS-01 and PR.PS-02 for the hostnames covered.
- The scope as tested and its exclusions. What was left out is evidence too: it shows where the report stops, and the third-party services excluded from testing can be compared with your supplier list under GV.SC-04.
The test window as a check on your own monitoring
Whether logs are kept and watched is a fact about how you operate, and a test of the website cannot show it. The test window does give you something to check. Afterwards, ask whoever runs your logs and alerts whether the tester's refused requests were recorded and whether any alert fired.
Their answer can be evidence for PR.PS-04 and DE.CM-09 that the test itself cannot produce. NIST's Implementation Examples for DE.CM-09 include "Monitor authentication attempts to identify attacks against credentials and unauthorized credential reuse".
Where a test stops: outcomes that need your own records
Outcomes about how the business is run cannot be evidenced by a test of one website or application.
The crosswalk marks those rows as organisational records rather than stretching a test result to cover them.
- Govern. Mission, stakeholder expectations, legal and contractual requirements, risk appetite, roles, policy, oversight and supply chain risk management are decisions and records of the business. A test can be scoped to answer a contract's testing clause, but it cannot show that the contract's requirements are "understood and managed", in the words of GV.OC-03.
- Awareness, training and physical access. Awareness and Training (PR.AT) concerns people, and PR.AA-06 reads "Physical access to assets is managed, monitored, and enforced commensurate with risk". Neither is visible from a browser.
- Backups and capacity. The outcome in PR.DS-11 is that backups of data are "created, protected, maintained, and tested", and your backup and restore records show that, not a test of the website. PR.IR-04 concerns the capacity to keep systems available. Load and denial-of-service testing sit outside every Clever Ops service, and that work needs a load-testing specialist and your host's written approval.
- Continuous monitoring. CSWP 29 says actions that support Govern, Identify, Protect and Detect should all happen continuously. A test lasts its agreed window. Clever Ops does not provide round-the-clock monitoring.
- Response and recovery. CSWP 29 says actions that support Respond and Recover should be ready at all times and happen when cybersecurity incidents occur. A test of the website does not exercise them. Clever Ops does not provide incident response or forensics. A site that is compromised now needs an incident response specialist first; after the cleanup, an audit can look at the web application before customers use it again.
- Everything outside the scope. Staff devices, office networks, email, cloud consoles, mobile apps and platforms that vendors run are outside a web application test, whatever its result.
How to use the crosswalk with a CSF request
Work from the request in front of you, not from the whole Core.
- List the Functions, Categories or Subcategories the contract, questionnaire or Target Profile names, with its exact wording. If it uses version 1.1 identifiers, translate them first, as set out below.
- Find each one in the crosswalk. Where a Subcategory is not there, read it in Appendix A of CSWP 29 and treat it as an organisational record unless your website or application plainly takes part in it.
- Put names in the owner column. The completed column is itself a record for GV.RR-02, which reads "Roles, responsibilities, and authorities related to cybersecurity risk management are established, communicated, understood, and enforced".
- Collect the artefact for each row and date it. Where a row says a test contributes, file the relevant report section next to the artefact, with the report's date and scope.
- Run the checks only on systems your business runs, with the limits written on each row. Four rows send requests to the website or application as a check: ID.AM-03, PR.AA-03, PR.AA-05 and PR.DS-02. In the rest you read, ask or write down.
- Answer each item from the evidence that covers it, and say where that evidence stops. A report on the customer portal does not answer a question about staff laptops.
If the request uses CSF 1.1 identifiers
An older questionnaire may quote version 1.1 identifiers. NIST's FAQ lists the major changes to the Core from version 1.1. Among them, in version 1.1 terms, much of Identify has moved to the new Govern Function, including ID.BE (now GV.OC), ID.RM (now GV.RM) and ID.SC (now GV.SC).
PR.AC (version 1.1) has been moved to PR.AA, revised and re-sequenced, and Identify has a new Improvement (ID.IM) Category. The FAQ says there was no removal of content from 1.1 to 2.0, that a Subcategory present in both versions has a similar meaning in both, and that PR.DS-3 from version 1.1 was moved to ID.AM-08 in CSF 2.0, with the identifier PR.DS-3 not reused. The FAQ also says the crosswalk between CSF 1.1 and CSF 2.0 is an Informative Reference.
In your answer, write the version 1.1 identifier the request used beside its CSF 2.0 home, so the reader can follow both.
Who decides whether the evidence is enough
NIST's FAQ says NIST does not offer certifications or endorsements of CSF-related products, implementations or services, and that there are no plans to develop a conformity assessment program.
So there is no CSF certificate issued by NIST for a business to hold. If someone offers one, ask who issues it.
The decision belongs to whoever asked. A customer that wrote the CSF into a contract, a parent company that set it as policy, or a board that adopted it decides what evidence satisfies it, and may use its own auditor or assessor to reach that view.
Questions about what a contract asks of you go to your adviser. A test report from any provider is evidence for that decision, not the decision itself.
If the request is for an ISO/IEC 27001 certificate or a SOC 2 report, a CSF crosswalk does not answer it, and the guides linked below cover each. Clever Ops is not an ISO certification body or a SOC 2 auditor.
For Australian readers comparing the CSF with local guidance, the Australian Signals Directorate (ASD) says it has developed prioritised mitigation strategies, the Strategies to mitigate cyber security incidents, to help organisations protect themselves against various cyber threats, and that the most effective of these are the Essential Eight. That makes the Essential Eight a set of mitigation strategies rather than an outcome taxonomy, and the Essential Eight guide covers it.
Where a Clever Ops test fits in a CSF programme
The systems Clever Ops tests are the ones a customer reaches with a browser or through an API.
The website security check looks from outside, without signing in, at certificates and transport security, headers, cookies, exposed files and admin paths, and the software versions a site shows to anyone. The WordPress security audit follows every role that can log in, from subscriber to administrator, and records the installed plugin and theme versions with evidence.
Behind a login, the web application and API penetration test of a custom application works through test accounts in each role, checking sign-in and sessions, the records and actions open to each role, whether one customer can get at another's records, and the API when it is called directly. Apps made with AI builders, and AI chatbots and agents, have checks of their own.
Each report sets out the scope as tested, every finding with its evidence and fix, the negative results and a remediation plan, and one consolidated retest is included once the fixes are deployed. Coverage is stated against the OWASP Web Security Testing Guide v4.2 and the OWASP Application Security Verification Standard 5.0.0.
Those references record which checks were run; they are not a certificate, and they give no opinion on compliance. In crosswalk terms, that evidence lands in the rows a test contributes to, chiefly ID.RA-01, ID.RA-06, ID.IM-02 and the Protect rows for the system tested.
A Clever Ops test is not a CSF assessment. It does not produce an Organizational Profile, choose a Tier, rate maturity or issue any certification, nor does Clever Ops hold any security certification or accreditation. Independence matters too: on a system Clever Ops built or hosts, the work is labelled a security review, and the report names the builder and the host, so it does not pass as an independent test.
When the party asking needs independence, choose a provider that had no hand in building or running the system.
Office networks, staff devices, phishing, load testing, incident response and monitoring go to other specialists. How a test runs, from the written scope and the authorisation letter through to the retest, is on the methodology page.
Frequently asked questions
This page makes no statement that any Australian law or regulator requires it. CSWP 29 says the CSF may be adopted voluntarily and through governmental policies and mandates. NIST's FAQ says NIST is not a regulatory agency, that Executive Order 13800 made the Framework mandatory for U.S. Federal Government agencies, and that some companies require the CSF for their customers or within their supply chain. If a customer contract or a parent company names the CSF, read its exact wording and take any question about what it asks of you to your adviser.
Not by NIST. Its FAQ says NIST does not offer certifications or endorsements of CSF-related products, implementations or services, and that there are no plans to develop a conformity assessment program. Whoever asked you to align with the CSF decides what evidence satisfies them. If you are offered a "NIST CSF certification", ask who issues it and what it examined.
The CSF does not provide one. NIST's FAQ says the CSF provides a series of outcomes to prioritise and address cybersecurity risks but does not specify actions for meeting them, and CSWP 29 says its outcomes are not a checklist of actions to perform. NIST's online Implementation Examples give notional steps towards the Subcategories' outcomes, and the publication says they are neither a comprehensive list of actions nor a baseline of required actions. The crosswalk on this page lists evidence for the website and application part only.
It can contribute evidence to a few Subcategories, such as ID.RA-01, on vulnerabilities being identified, validated and recorded, and ID.IM-02, on improvements identified from security tests and exercises, for which NIST's Implementation Examples include penetration testing of selected high-risk systems as approved by leadership. It cannot evidence governance, training, backups, monitoring, incident response or recovery, and it covers only the system, scope and dates tested. Whether the evidence is enough is for whoever asked you to align with the CSF to decide.
Translate each one to its CSF 2.0 home and answer against that. NIST's FAQ says PR.AC (version 1.1) has been moved to PR.AA, revised and re-sequenced, that ID.SC (version 1.1) is now GV.SC, and that the meaning of all 1.1 content is included within CSF 2.0. The FAQ also says the crosswalk between CSF 1.1 and CSF 2.0 is an Informative Reference. Give both identifiers in your answer, so the customer can match it to their question.
Scope the right test for your application
Tell us what you have built, who uses it and when it needs to launch. You will get a plain-English recommendation: a website audit, a web application test or a referral to a specialist.