Skip to main content
Web securityReviewed

Vulnerability assessment for websites and web applications: each finding confirmed, ranked and assigned

A broad search of your website or web application for known kinds of weakness, from the open internet and from inside each signed-in role.

A tester confirms each finding before it is reported. It is then ranked by what it exposes, who can reach it and whether attackers are known to use it, and it reaches you with an owner, a fix and the check that will show the fix worked.

What is a vulnerability assessment?

A vulnerability assessment looks across a system for weaknesses, confirms which ones are real and ranks them, so the people responsible know what to fix first.

The definition NIST uses in SP 800-53 Rev. 5, taken from CNSSI 4009-2022, is a wide one: a systematic examination of an information system or product to determine the adequacy of security measures, identify security deficiencies, provide data from which to predict the effectiveness of proposed security measures, and confirm the adequacy of such measures after implementation.

Two words in that definition are worth holding a provider to. Ask what systematic looked like: was the coverage planned and written down, so you can see what was examined as well as what was found? Then ask about confirm, in the last clause: once the list is delivered, who checks that the fixes held?

The same NIST glossary defines penetration testing, for SP 800-53 Rev. 5, as a test methodology in which assessors, typically working under specific constraints, attempt to circumvent or defeat the security features of a system. The difference is direction. An assessment goes wide, finding and ranking known kinds of weakness one at a time.

A penetration test goes deep, trying to defeat the controls that matter most and seeing how far that leads. Clever Ops offers both for web applications. This page covers the assessment.

For a web application, the weaknesses in scope are the well-documented ones: missing access checks between accounts and between roles, injection into queries and pages, weak login and session handling, components with published advisories, exposed files, and misconfigured headers, cookies and cross-origin rules.

An assessment suits an application that has never been looked at, one that has changed a lot since its last test, or a team that wants a ranked backlog before deciding whether a deeper test is worth it.

Why a list of scan results is not an assessment

Some work sold as a vulnerability assessment is an automated scan with a cover page.

The output is long, sorted by the scanner's own severity labels, and full of items nobody has checked. NIST's Technical Guide to Information Security Testing and Assessment, SP 800-115, sets out why that is not enough. As with all pattern-matching and signature-based tools, application vulnerability scanners typically have high false positive rates, and they also suffer from high false negative rates.

Scanners often define risk levels by their own proprietary methods, which makes findings from different scanners hard to compare.

The guide's sharper point is about severity. The risk level a scanner assigns may not reflect the actual risk to the organisation. Its example is an FTP server flagged as a moderate risk because it sends passwords in clear text, when the organisation runs it only as an anonymous public server that uses no passwords at all.

SP 800-115 says assessors should determine the appropriate risk level for each vulnerability rather than simply accept the levels a scanner assigns.

Web findings work the same way: an outdated script library that loads only on a public brochure page is a different problem from the same library running inside the staff dashboard.

Some weaknesses never show up in scan output. The OWASP Web Security Testing Guide v4.2 says business logic flaws cannot be detected by a vulnerability scanner. A sound assessment report therefore states which kinds of weakness it covered and which it left for a penetration test, instead of letting a long list imply completeness.

At Clever Ops, each finding is validated by a tester before it is reported, and prioritised by the asset it affects, the roles that can reach it and its exploit context. A result that could not be reproduced stays out of the findings and is recorded with the checks that found nothing. Clever Ops does not sell raw scan output, or a scan subscription, as a service of its own.

Four triage rules that turn results into a work list

The sample triage queue on this page applies four rules in order.

They work on any list of security results, whether it came from a scan you ran, an earlier report or an email from a researcher, so you can use them before you buy anything.

Merge by root cause and owner, not by label

Two results are one finding when one change fixes both and one person owns that change. A missing ownership check inside a shared lookup function can surface on four routes and still be a single finding with four affected routes. Two results with the same label stay separate when they live in different code bases with different owners, such as script injection in the marketing site search and in the customer portal search.

Merging by label hides work, and splitting by URL multiplies it: in the sample, 225 raw results become nine queue items.

Confirm with the smallest proof that settles it

Validation means reproducing a result with the least intrusive request that shows it is real, on test accounts and test records, and keeping the request and response as evidence. A component is confirmed by its exact version and by whether the affected feature is switched on.

Access between customers is confirmed the way WSTG-ATHZ-02 describes horizontal testing: two users holding identical privileges, both signed in, with the parameters and session identifier of one substituted into the other's requests.

A result that cannot be checked with the access in scope becomes a listed limitation, with a note of what would settle it.

Rank by asset, role and exploit context

Priority comes from three questions asked of every confirmed finding. What does the affected asset hold or control: customer records, money, staff functions, or a public page with nothing behind it? Which roles can reach it: anyone on the internet, anyone who can register, a role your team issues, or staff only?

What is the exploit context: is the weakness in a published component that attackers are known to use, does it need a victim to click something, or does it depend on a condition the assessment did not find?

A finding that anyone can reach and that exposes customer data goes to the top, whatever its label says.

Hand each finding to an owner who can act alone

A finding is ready to hand over when one person can start on it without a meeting. That takes an owner named by role, such as the portal developer, the site maintainer or the hosting provider, plus the affected routes, the redacted evidence, a fix written for your stack, and retest criteria stating what the second look will check.

Naming the owner by role keeps the list usable when people change jobs or a contractor takes over.

Severity scores, known-exploited lists and exploit likelihood

Three public sources are commonly used to rank vulnerabilities, and each one answers a different question.

Read together they are useful. Read alone, each misleads in a predictable direction.

CVSS v3.1: how severe a weakness is in general

The FIRST specification for CVSS v3.1 builds a score from Base, Temporal and Environmental metric groups, and the Base score runs from 0.0 to 10.0. Base scores are usually produced by the organisation that maintains the vulnerable product, or a third party scoring on its behalf, and the specification says consumers should supplement them with Temporal and Environmental scores specific to their own use of the product.

The Environmental metrics let an analyst adjust the score for the importance of the affected IT asset to the organisation. A published score is a starting point, not your priority.

The sample queue shows the gap. Its top finding, one customer reading another customer's invoices, scores 6.5 on the CVSS v3.1 Base metrics, which the specification's scale calls Medium (vector AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N). It still heads the queue, because anyone can register as a customer and the records are financial. The report carries both numbers for each rated finding: the CVSS score with its vector, and a business-risk rating.

The KEV catalogue: whether attackers are known to use it

The US Cybersecurity and Infrastructure Security Agency (CISA) describes its Known Exploited Vulnerabilities (KEV) catalogue as the authoritative source of vulnerabilities that have been exploited in the wild, and says organisations should use it as an input to their vulnerability management prioritisation framework. When a plugin, library or server component in your application carries a vulnerability on that list, the exploit context question has its answer.

A weakness in code only your business runs is unlikely to have a public advisory at all, so the catalogue speaks to the components you use rather than to your own code.

EPSS: how likely exploitation is in the near term

The Exploit Prediction Scoring System (EPSS), also from FIRST, estimates the probability that a published CVE will be exploited in the wild in the next 30 days, and publishes a score between 0 and 1 for every CVE each day.

FIRST's own guidance on using it is frank about the limits: EPSS does not know what is in your environment, it has no visibility into your asset inventory or compensating controls, and it is not a severity score or a complete risk score.

As a general rule of thumb, the same guidance says to treat a vulnerability that appears on CISA's KEV list as actively exploited and prioritise it accordingly, regardless of its EPSS score.

That guidance also locates the work an assessment does. It says remediation effort produces the most risk reduction where a high EPSS score meets confirmed presence, confirmed reachability and high consequence. Presence and reachability are what validation establishes, and consequence is what the asset and role questions answer.

External and authenticated coverage

An external vulnerability assessment of a website covers what anyone on the internet can reach without signing in: public pages and forms, sign-up, login and password reset, public API routes, the code your site sends to each browser, and the components and versions that code gives away.

For a brochure site where only your own staff sign in, that is most of the exposure, and the website security audit works through it check by check.

In this service, external means the website or web application as the internet sees it. It does not reach your network perimeter: firewalls, VPN gateways, mail servers, remote access services and the other hosts an external network vulnerability assessment would include. Those belong with a network security specialist.

Authenticated coverage begins once test accounts are in place. With two accounts for each role in scope, plus accounts in two customer organisations if your application serves more than one, the assessment checks what each role can reach, whether one account can read or change another account's records, and which components and pages load only after sign-in.

If your customers hold accounts, this half usually carries the findings that matter most, because it is where one customer's data sits beside another's.

Each finding records the view it was seen from. One marked unauthenticated is reachable by anyone; one marked authenticated names the role it needed. Gaps are recorded the same way, so a role with no test account appears as not covered instead of disappearing from the report.

  • Signed out:public pages and forms, sign-up, login and reset flows, public API routes, browser code and the versions it reveals.
  • Signed in:each role's pages and API routes, access between accounts, between roles and between customer organisations, components loaded after sign-in, file uploads, and stored input that other users later see.
  • Neither view:source code unless code access is added, the business rules your application is meant to enforce, services run by other businesses, and your office network.

Sample triage queue: 225 raw results to nine owner-ready items

An illustration written for this page, not a client's results. orders.lab.test is an imaginary ordering portal with open customer sign-up, customer users, customer administrators and staff, and www.lab.test is its marketing site. To reuse the queue, put your own results in the first column and apply four rules in order.

How to use it

Merge: results are one item when one change fixes them and one person owns that change. Validate: confirm with the least intrusive proof, or mark the item closed (not reproduced) or unrated (not checkable with the access in scope). Prioritise: P1 when anyone, or anyone who can register, can reach it and it exposes customer data, money or control of the site, or when it is on the KEV catalogue and reachable; P2 when it needs a role you issue or a victim's click, and exposes data or functions beyond that role; P3 when the impact is limited or depends on a condition not present; P4 for hardening. Hand off: an owner by role, a fix and retest criteria. The two coverage columns show whether each item was checked signed out, signed in or not at all. Every hostname, account, advisory status and result below is invented.

9 rows

Sample triage queue: 225 raw results to nine owner-ready items

Queue item and raw results mergedAsset and what it holdsUnauthenticated coverageAuthenticated coverageValidation resultWho can reach it and exploit contextPriority and whyOwner-ready handoff
Q1. Four raw results merged: another account's record returned on /api/invoices/{id}, /api/invoices/{id}/pdf, /api/statements/{id} and /api/quotes/{id}. One shared lookup function serves all four routes, so it is one finding.orders.lab.test customer API. Invoices, statements and quotes: customer names, delivery addresses, order values and account balances.Checked. All four routes return 401 without a session, so a stranger sees nothing.Checked with cust-a1 and cust-b1, customer users in two organisations. cust-b1's session read cust-a1's records on all four routes. Changes were not attempted.Confirmed. Four redacted request pairs, each shown beside the owner's own request for the same record.Any signed-in customer. Sign-up is open and record ids are sequential, so anyone can register and step through them. Custom code, so no CVE, KEV entry or EPSS score applies.P1. Anyone can reach it by registering, and it exposes other businesses' financial records. CVSS v3.1 Base 6.5, Medium; the business rating moves it to the top.Owner: portal API developer. Fix: compare the record's organisation with the session's organisation inside the shared lookup, so all four routes inherit the check. Retest: cust-b1 receives 403 or 404 on each route, and cust-a1 still reads its own records.
Q2. Two raw results merged: an outdated form plugin reported on the contact page and again on the quote page. One installation, one update.www.lab.test marketing site on a CMS. Enquiry forms, and an admin area that can publish to the public site.Checked. The plugin version shows in public asset paths, and the endpoint named in the vendor's advisory answers requests that carry no session.Not needed: the affected endpoint requires no sign-in. CMS admin roles were outside the scope.Confirmed by exact version and a reachable endpoint. Exploitation was not attempted, under the non-destructive defaults.Anyone on the internet. In this illustration the advisory's vulnerability is on CISA's KEV catalogue, so attackers are known to use it.P1. Known exploited and reachable without an account. The low value of a brochure site does not lower it, because control of the site is the impact.Owner: site maintainer. Fix: update the plugin to the vendor's fixed release ahead of the rest of the queue, then look for signs of compromise; if there are any, call an incident response specialist first. Retest: installed version at or above the fixed release.
Q3. One raw result: a staff-only route, /api/admin/price-lists, named in the dashboard script served to customer administrators.orders.lab.test staff API. Cost prices and supplier margins for every product.Checked. 401 without a session.Checked with cadmin-a, a customer administrator, and cust-a1, a customer user. cadmin-a received the full price list; cust-a1 received 403.Confirmed. Your team confirmed in writing that customer administrators should not see cost prices, so the broken rule is stated, not guessed.Customer administrators only, one or two per customer organisation, created by your staff. No published advisory applies.P2. Commercially sensitive data, but reachable only through a role you issue. Sits below Q1 because fewer, known people hold that role.Owner: portal API developer. Fix: enforce the staff role on the server for every /api/admin route, not only this one. Retest: cadmin-a receives 403 on this route and on the other admin routes listed in the evidence.
Q4. One raw result: text typed into the order search box comes back unencoded on the results page.orders.lab.test order history. The signed-in customer's own orders and delivery addresses.Not reachable. The search page requires a session.Checked with cust-a1. A harmless marker element, sent in a link that cust-a1 opened, rendered as markup on the signed-in results page.Confirmed. Content carried in a link renders inside the victim's own signed-in page.An outsider who persuades a signed-in customer to open a crafted link. Needs the victim to click. No published advisory applies.P2. Runs within a customer's session and could read that customer's orders. Needs a click, so it sits below Q1 and Q2.Owner: portal front-end developer. Fix: encode search terms on output, and add a Content Security Policy to the portal. Retest: the marker renders as plain text on the results page and in the page title.
Q5. One raw result with the same label as Q4: the marketing site search echoes its query unencoded. Kept apart from Q4 because a different code base and a different owner fix it.www.lab.test public search. No accounts and no session cookies.Checked. The marker rendered as markup on the public results page.Not applicable. The marketing site has no customer sign-in.Confirmed.Anyone who can get a visitor to open a crafted link. With no session to steal, the realistic misuse is showing a visitor false content on your domain.P3. Same weakness as Q4, lower priority, because nothing on the page belongs to a signed-in user.Owner: site maintainer. Fix: encode the query on output in the search template. Retest: the marker renders as plain text.
Q6. Two raw results merged: an outdated rich-text editor on the customer notes page and in the staff ticket view. Both load the same bundled file.orders.lab.test notes and tickets. Text that customers write and staff open later.Not reachable. The editor loads only after sign-in.Checked with cust-a1, where the version shows in the script served after sign-in. The staff view could not be opened, because no staff account was in scope.Confirmed by version only. In this illustration the vendor's advisory needs a paste option that this configuration leaves switched off, so exploitation here was not shown.Any signed-in customer, aiming at staff who open their notes. In this illustration the vulnerability is not on the KEV catalogue and has a low EPSS score.P3. Rises to P2 if the paste option is switched on, because staff sessions would then be exposed.Owner: portal front-end developer. Fix: update the editor in the next release, and record that the paste option must stay off until then. Retest: the served version is at or above the vendor's fixed release.
Q7. 212 raw results merged: the same web server version header on every response from both hostnames. One setting at the hosting layer.www.lab.test and orders.lab.test responses. No data; the header names the web server software and its version.Checked on both hostnames.Checked. The same header appears on signed-in responses.Confirmed, and reported once rather than 212 times.Anyone. It tells an attacker what to look up but is not a weakness on its own. In this illustration the version shown has no open advisory.P4. Hardening. Fix it the next time the hosting settings change.Owner: hosting provider, or whoever manages the server settings. Fix: remove the version from the header. Retest: no version string in the response headers on either hostname.
Q8. One raw result: possible SQL injection in the product filter, based on one slower response.orders.lab.test product catalogue. Public product data, plus customer-specific prices after sign-in.Checked. The filter is public.Checked again with cust-a1, where the filter also returns that customer's prices.Not reproduced. Repeated paired requests and syntax-breaking inputs showed no consistent timing difference, and errors stayed generic. Closed as a false positive.Not applicable once closed.Closed. Not a finding. Listed with the checks that found nothing, along with the requests tried.Owner: none. Your developer can see what was tried, and the retest does not revisit it.
Q9. One raw result: the staff route that imports a supplier catalogue from a web address accepts any address, as seen in the dashboard script.orders.lab.test staff tools. The server fetches whatever web address it is given.Checked. 401 without a session.Not covered. No staff test account was in scope, and customer roles receive 403.Not validated. Listed as a limitation, with what would settle it: a staff test account and a harmless address the tester controls.Staff only, so an outsider would need a staff account first.Unrated. A limitation, not a finding, until it is validated.Owner: you, to decide whether to add a staff account. That is new scope, agreed separately, because the retest covers reported findings only.

Queue item and raw results merged

Asset and what it holds

  1. Q1. Four raw results merged: another account's record returned on /api/invoices/{id}, /api/invoices/{id}/pdf, /api/statements/{id} and /api/quotes/{id}. One shared lookup function serves all four routes, so it is one finding.
    Asset and what it holds
    orders.lab.test customer API. Invoices, statements and quotes: customer names, delivery addresses, order values and account balances.
    Unauthenticated coverage
    Checked. All four routes return 401 without a session, so a stranger sees nothing.
    Authenticated coverage
    Checked with cust-a1 and cust-b1, customer users in two organisations. cust-b1's session read cust-a1's records on all four routes. Changes were not attempted.
    Validation result
    Confirmed. Four redacted request pairs, each shown beside the owner's own request for the same record.
    Who can reach it and exploit context
    Any signed-in customer. Sign-up is open and record ids are sequential, so anyone can register and step through them. Custom code, so no CVE, KEV entry or EPSS score applies.
    Priority and why
    P1. Anyone can reach it by registering, and it exposes other businesses' financial records. CVSS v3.1 Base 6.5, Medium; the business rating moves it to the top.
    Owner-ready handoff
    Owner: portal API developer. Fix: compare the record's organisation with the session's organisation inside the shared lookup, so all four routes inherit the check. Retest: cust-b1 receives 403 or 404 on each route, and cust-a1 still reads its own records.
  2. Q2. Two raw results merged: an outdated form plugin reported on the contact page and again on the quote page. One installation, one update.
    Asset and what it holds
    www.lab.test marketing site on a CMS. Enquiry forms, and an admin area that can publish to the public site.
    Unauthenticated coverage
    Checked. The plugin version shows in public asset paths, and the endpoint named in the vendor's advisory answers requests that carry no session.
    Authenticated coverage
    Not needed: the affected endpoint requires no sign-in. CMS admin roles were outside the scope.
    Validation result
    Confirmed by exact version and a reachable endpoint. Exploitation was not attempted, under the non-destructive defaults.
    Who can reach it and exploit context
    Anyone on the internet. In this illustration the advisory's vulnerability is on CISA's KEV catalogue, so attackers are known to use it.
    Priority and why
    P1. Known exploited and reachable without an account. The low value of a brochure site does not lower it, because control of the site is the impact.
    Owner-ready handoff
    Owner: site maintainer. Fix: update the plugin to the vendor's fixed release ahead of the rest of the queue, then look for signs of compromise; if there are any, call an incident response specialist first. Retest: installed version at or above the fixed release.
  3. Q3. One raw result: a staff-only route, /api/admin/price-lists, named in the dashboard script served to customer administrators.
    Asset and what it holds
    orders.lab.test staff API. Cost prices and supplier margins for every product.
    Unauthenticated coverage
    Checked. 401 without a session.
    Authenticated coverage
    Checked with cadmin-a, a customer administrator, and cust-a1, a customer user. cadmin-a received the full price list; cust-a1 received 403.
    Validation result
    Confirmed. Your team confirmed in writing that customer administrators should not see cost prices, so the broken rule is stated, not guessed.
    Who can reach it and exploit context
    Customer administrators only, one or two per customer organisation, created by your staff. No published advisory applies.
    Priority and why
    P2. Commercially sensitive data, but reachable only through a role you issue. Sits below Q1 because fewer, known people hold that role.
    Owner-ready handoff
    Owner: portal API developer. Fix: enforce the staff role on the server for every /api/admin route, not only this one. Retest: cadmin-a receives 403 on this route and on the other admin routes listed in the evidence.
  4. Q4. One raw result: text typed into the order search box comes back unencoded on the results page.
    Asset and what it holds
    orders.lab.test order history. The signed-in customer's own orders and delivery addresses.
    Unauthenticated coverage
    Not reachable. The search page requires a session.
    Authenticated coverage
    Checked with cust-a1. A harmless marker element, sent in a link that cust-a1 opened, rendered as markup on the signed-in results page.
    Validation result
    Confirmed. Content carried in a link renders inside the victim's own signed-in page.
    Who can reach it and exploit context
    An outsider who persuades a signed-in customer to open a crafted link. Needs the victim to click. No published advisory applies.
    Priority and why
    P2. Runs within a customer's session and could read that customer's orders. Needs a click, so it sits below Q1 and Q2.
    Owner-ready handoff
    Owner: portal front-end developer. Fix: encode search terms on output, and add a Content Security Policy to the portal. Retest: the marker renders as plain text on the results page and in the page title.
  5. Q5. One raw result with the same label as Q4: the marketing site search echoes its query unencoded. Kept apart from Q4 because a different code base and a different owner fix it.
    Asset and what it holds
    www.lab.test public search. No accounts and no session cookies.
    Unauthenticated coverage
    Checked. The marker rendered as markup on the public results page.
    Authenticated coverage
    Not applicable. The marketing site has no customer sign-in.
    Validation result
    Confirmed.
    Who can reach it and exploit context
    Anyone who can get a visitor to open a crafted link. With no session to steal, the realistic misuse is showing a visitor false content on your domain.
    Priority and why
    P3. Same weakness as Q4, lower priority, because nothing on the page belongs to a signed-in user.
    Owner-ready handoff
    Owner: site maintainer. Fix: encode the query on output in the search template. Retest: the marker renders as plain text.
  6. Q6. Two raw results merged: an outdated rich-text editor on the customer notes page and in the staff ticket view. Both load the same bundled file.
    Asset and what it holds
    orders.lab.test notes and tickets. Text that customers write and staff open later.
    Unauthenticated coverage
    Not reachable. The editor loads only after sign-in.
    Authenticated coverage
    Checked with cust-a1, where the version shows in the script served after sign-in. The staff view could not be opened, because no staff account was in scope.
    Validation result
    Confirmed by version only. In this illustration the vendor's advisory needs a paste option that this configuration leaves switched off, so exploitation here was not shown.
    Who can reach it and exploit context
    Any signed-in customer, aiming at staff who open their notes. In this illustration the vulnerability is not on the KEV catalogue and has a low EPSS score.
    Priority and why
    P3. Rises to P2 if the paste option is switched on, because staff sessions would then be exposed.
    Owner-ready handoff
    Owner: portal front-end developer. Fix: update the editor in the next release, and record that the paste option must stay off until then. Retest: the served version is at or above the vendor's fixed release.
  7. Q7. 212 raw results merged: the same web server version header on every response from both hostnames. One setting at the hosting layer.
    Asset and what it holds
    www.lab.test and orders.lab.test responses. No data; the header names the web server software and its version.
    Unauthenticated coverage
    Checked on both hostnames.
    Authenticated coverage
    Checked. The same header appears on signed-in responses.
    Validation result
    Confirmed, and reported once rather than 212 times.
    Who can reach it and exploit context
    Anyone. It tells an attacker what to look up but is not a weakness on its own. In this illustration the version shown has no open advisory.
    Priority and why
    P4. Hardening. Fix it the next time the hosting settings change.
    Owner-ready handoff
    Owner: hosting provider, or whoever manages the server settings. Fix: remove the version from the header. Retest: no version string in the response headers on either hostname.
  8. Q8. One raw result: possible SQL injection in the product filter, based on one slower response.
    Asset and what it holds
    orders.lab.test product catalogue. Public product data, plus customer-specific prices after sign-in.
    Unauthenticated coverage
    Checked. The filter is public.
    Authenticated coverage
    Checked again with cust-a1, where the filter also returns that customer's prices.
    Validation result
    Not reproduced. Repeated paired requests and syntax-breaking inputs showed no consistent timing difference, and errors stayed generic. Closed as a false positive.
    Who can reach it and exploit context
    Not applicable once closed.
    Priority and why
    Closed. Not a finding. Listed with the checks that found nothing, along with the requests tried.
    Owner-ready handoff
    Owner: none. Your developer can see what was tried, and the retest does not revisit it.
  9. Q9. One raw result: the staff route that imports a supplier catalogue from a web address accepts any address, as seen in the dashboard script.
    Asset and what it holds
    orders.lab.test staff tools. The server fetches whatever web address it is given.
    Unauthenticated coverage
    Checked. 401 without a session.
    Authenticated coverage
    Not covered. No staff test account was in scope, and customer roles receive 403.
    Validation result
    Not validated. Listed as a limitation, with what would settle it: a staff test account and a harmless address the tester controls.
    Who can reach it and exploit context
    Staff only, so an outsider would need a staff account first.
    Priority and why
    Unrated. A limitation, not a finding, until it is validated.
    Owner-ready handoff
    Owner: you, to decide whether to add a staff account. That is new scope, agreed separately, because the retest covers reported findings only.

What the assessment delivers

The report is built to be worked through rather than filed.

Its centre is the ranked queue, and everything else in it exists so a customer, auditor or your own developer can check how the queue was built.

  • The scope as assessed: hostnames, API addresses, roles and test accounts, the build, the dates, and anything excluded.
  • A coverage map that shows, for each area of the application, whether it was assessed signed out, signed in or not at all, and why.
  • The findings in priority order, duplicates merged, with every affected route listed under the finding it belongs to.
  • For each finding: redacted evidence, the roles that can reach it, its exploit context, a CVSS v3.1 score with its vector, a business-risk rating, a fix written for your stack, an owner by role and retest criteria.
  • Each finding mapped to its OWASP Top 10:2025 category, with coverage referenced to the OWASP Web Security Testing Guide v4.2 and ASVS 5.0. That records what was checked; it is not a certificate.
  • Checks that found nothing, recorded against the build and accounts used, so a later question about whether something was looked at has an answer.
  • Results that could not be confirmed, listed as limitations alongside what would settle each one.
  • One consolidated retest of the reported findings, after which each finding carries one of four statuses: fixed, partially fixed, still reproducible or not retested.

How the retest works

You request the retest within the window set in your statement of work, once your fixes are deployed to an identified build of the same application. It covers the reported findings within the agreed scope, and new features or unrelated changes are scoped as separate work.

A finding left out of the retest is shown as not retested rather than as fixed, so the updated report says no more than was checked. Your own developers, your agency or a separately agreed Clever Ops security fix sprint can do the fixing.

Where an assessment stops and a penetration test starts

An assessment ranks weaknesses one at a time, and SP 800-115 names the catch.

Vulnerabilities rarely exist in isolation: several low-risk ones can present a higher risk when combined, and a tool may give each a low rating and leave the assessor falsely confident. The guide calls penetration testing a more reliable way to identify the risk of vulnerabilities in aggregate.

The NIST glossary entry drawn from the same publication adds that most penetration tests look for combinations of vulnerabilities that give more access than any single one would.

So the assessment notes combinations it can see, such as an error message that leaks the record ids an access gap needs, but it does not chain weaknesses to find out how far they lead, and it does not set out to break the business rules your application enforces. Choose a web application penetration test instead of the assessment, or after it, when:

  • Customers pay, book, claim credit or redeem codes inside the application, so a broken rule would cost money.
  • One database serves several customer organisations, and the boundary between them is the thing you most need tested.
  • A customer, insurer or contract asks for a penetration test by name, rather than for an assessment.
  • The assessment left questions only a deeper look can settle, such as a weakness whose impact depends on another.
  • You need to know how far an attacker could get, not only what is wrong.

The queue becomes the scope

If both are on the table, the assessment's queue is a strong input to a penetration test scope. It shows which routes, roles and components already carried confirmed weaknesses, so the deeper work starts where the risk has already shown itself instead of repeating the broad pass.

How the assessment behaves on a live application

Validation sends real requests to a real application, so the assessment runs under the same rules as any Clever Ops test.

Each target is named in a signed authorisation from someone entitled to approve testing, and your hosting provider's testing policy is checked for each target before testing begins.

  • Destructive actions are off by default for real people's accounts and for records that existed before the assessment. Reset, lockout and deletion checks run only on test accounts and test records, where the rules of engagement allow them.
  • Each finding is confirmed with the least intrusive proof available. A key found in browser code is not tried, and a backup file found on the server is not downloaded in full.
  • Where a staging copy matches production, the assessment starts there. On production, the request rate is capped, and testing on a target pauses if the application or its firewall starts refusing requests.
  • Test data is tracked as it is created, then removed or listed in the report for your team to remove.
  • Evidence keeps only what a finding needs, with tokens, passwords and personal details redacted. It is stored encrypted in Australia and deleted on the schedule in the data handling terms agreed before testing.
  • Denial-of-service, load and volumetric testing are outside the assessment.

Vulnerability assessment services in Australia: fit and referrals

Clever Ops is a small Australian consultancy based in Gippsland, Victoria, working with businesses earning $1M to $50M in annual revenue anywhere in Australia.

The assessment covers websites, WordPress and WooCommerce sites, and custom web applications with their APIs. Some requests that arrive as a vulnerability assessment need a different provider, and it is better to say so on the first call than after a scoping round.

  • A PCI DSS external scan. The PCI Security Standards Council says an Approved Scanning Vendor conducts external vulnerability scanning to validate adherence with the external scanning requirements of PCI DSS Requirement 11.3.2. Clever Ops is not a PCI Approved Scanning Vendor, so a merchant who needs those scans books them with one; this assessment is evidence about the application, not a substitute for them.
  • A tender, panel or contract that requires a CREST-accredited provider. CREST lists Vulnerability Assessment among the disciplines its accreditation covers, alongside Penetration Testing. Clever Ops is not CREST accredited, so we refer that work to a provider that is.
  • A network vulnerability assessment of servers, firewalls, VPNs, office networks, staff devices or Active Directory. That work goes to a network security specialist.
  • A site that is compromised now. Containment and cleanup by an incident response specialist come first, and an assessment can follow once the site is clean.
  • A certificate or compliance sign-off. An assessment report is evidence for a decision someone else makes, such as an auditor or a customer, and it does not certify a site.
  • Live Shopify storefronts and Laravel applications, which sit outside this service.
  • A website or application Clever Ops built or hosts. The same work then runs as a security review rather than an independent assessment, and the report states who built and who hosts the system.

Frequently asked questions

A structured search of a system for weaknesses, followed by confirming and ranking what turned up. The NIST glossary definition used in SP 800-53 Rev. 5 calls it a systematic examination to determine the adequacy of security measures and identify security deficiencies. For a website or web application, that means covering the known kinds of weakness signed out and signed in, confirming each result, and handing over a list in the order it should be fixed.

A scan compares what it sees with known patterns and reports possible problems. NIST SP 800-115 notes that application vulnerability scanners typically have high false positive rates and that the risk level a scanner assigns may not reflect the actual risk to the organisation. An assessment does the sorting a scan leaves to you: duplicates merged, results confirmed, priorities set from your own assets and roles, and an owner and fix for each finding.

No. An assessment goes wide and ranks known kinds of weakness one by one. A penetration test, in the NIST definition, is an attempt to circumvent or defeat a system's security features, so it goes deeper: chaining weaknesses, testing business rules and showing how far an attacker could get. Many applications benefit from the assessment first, with its queue shaping the scope of a later test.

For the authenticated half, yes: two accounts for each role in scope, and accounts in two customer organisations if your application serves more than one. Without them the assessment covers only what a stranger can reach, and the report lists the signed-in areas as not covered rather than implying they were checked.

No. Clever Ops is not a PCI Approved Scanning Vendor. If your acquirer or payment provider asks for ASV scans, book them with a vendor on the PCI Security Standards Council's list. A web application assessment can sit alongside those scans as evidence about the application itself.

It does not become a finding. If it could not be reproduced, it is recorded among the checks that found nothing, together with the requests that were tried. If it could not be checked with the access in scope, such as a feature only a role without a test account can reach, it is listed as a limitation with what would settle it, and you decide whether to add that access as separate work.

Each component is identified by its exact version where the application reveals it, matched to the vendor's own advisories, and checked for whether the affected feature is in use. Exploit context comes from public sources: whether the vulnerability is in CISA's KEV catalogue, and its EPSS score where it has a CVE. A version match on its own is reported with that limit stated, not presented as a demonstrated exploit.

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.