Web application penetration testing: what is tested, and how the scope is built
A signed-in test of your custom web application and the API behind it, run from test accounts in every role.
This page sets out what a web app penetration test examines, how your routes, roles and business rules become test cases, and a worked scope for an invented twenty-route application, exclusions included.
On this page
What web application penetration testing covers
Web application penetration testing is a signed-in test of software that customers or staff log in to: a customer portal, a booking or quoting product, a software product sold by subscription, or an internal tool that can be reached from the internet.
The tester holds test accounts in each role the scope names, works through the application and the API behind it, and tries to make it do what its owners did not intend: show one customer another customer's records, let an ordinary user reach a staff function, mark an unpaid order as paid, or keep a file the server will run as code. Attempts that work become findings with evidence. Attempts that fail are recorded too.
Clever Ops tests custom web applications and their APIs, such as applications built on Next.js, React or Node. The penetration testing services hub compares this test with the lighter website security check, which stays outside the login, and with the other services we offer. This page goes one level down: what a web app penetration test examines inside an application, and how a written scope decides which of those things are tested in yours.
The methodology guide covers the other half, which is how each test is run, evidenced, rated and retested.
Providers sell similar work as application penetration testing, a web application security assessment or an application security test. The name tells you less than three answers you can get before you book: which roles the tester will hold accounts in, whether the API will be called directly or only through the screens, and which features the scope lists as excluded.
The four lists a web application scope is built from
A scope that names only a hostname leaves the tester to guess what matters, and leaves you unable to check afterwards what was covered.
For a web application, Clever Ops builds the scope from four lists, prepared with your developer before testing begins. Each list answers one question, and the test cases come from setting the lists against each other.
A route here means a page or API endpoint that accepts input or returns data that is not public. For each one, the list records the method and path, the roles allowed to call it, whether it reads or changes data, the kind of record it points at, and whether it is one step of a longer flow such as checkout.
That mirrors where WSTG v4.2 starts: its entry point test notes the parameters of each request, whether the request is authenticated, and whether it belongs to a multi-step process (WSTG-INFO-06).
Take the list from the routing files and any API schema rather than from clicking through the screens, because the screens show only the routes someone chose to link. Routes the front end has stopped calling belong on it as well. OWASP's API Security Top 10 says attackers usually get unauthorised access through old API versions or endpoints left running unpatched (API9:2023).
The second list is a grid with a row for each function and a column for each role, every cell marked allowed or refused. Writing it down tends to raise questions before any testing starts, such as whether a support role should be able to issue credit, or whether every role should be able to export the customer list.
WSTG v4.2's role definitions test asks the tester to identify and document the roles, try to switch to or reach another role, and review the granularity of the roles and the needs behind the permissions each one is given (WSTG-IDNT-01).
In the scope, every refused cell becomes at least one test case, sent from the role that should be refused.
The third list names the kinds of record the application keeps, such as customers, quotes, invoices and uploaded files, and who owns each kind. Ownership sets up the horizontal checks, where two accounts in the same role try to reach each other's records. WSTG v4.2 begins its insecure direct object reference test by mapping every place where user input refers directly to a database row, a file or a page (WSTG-ATHZ-04).
Files belong on the list: a PDF generated for one customer is as private as the row it was built from.
The last list holds the rules the server is meant to enforce and the screens only suggest: plan limits, discount caps, steps that must happen in order, actions that may happen once. WSTG v4.2 notes that a third-party tester has to ask the business whether different operations should be allowed by the application.
A short list in plain sentences, such as "a free account can send five quotes a month", gives each logic test a rule to measure against.
Set the four lists against each other and the test cases follow. Each route is tried from each role's accounts, against records someone else owns, under the rules that apply to it. That is why a modest application produces a long test plan, and why the worked scope on this page needs more than twenty test cases for twenty routes and two roles.
What a web app penetration test examines, area by area
These are the areas the worked scope covers, each with what is tried and what a control that holds looks like.
Where a WSTG v4.2 test or an ASVS 5.0 requirement covers the check, its id follows in brackets. The methodology guide explains how evidence is gathered for each finding. This section is about what the tester puts in front of the application.
Authentication is a set of flows rather than one login form. Sign-up, sign-in and password reset are compared for an email address that has an account and one that does not, because a different message, status code or response time lets an outsider build a list of your users (WSTG-IDNT-04; ASVS v5.0.0-6.3.8).
Sign-in is checked for a lockout after repeated wrong passwords, and for how a locked account is unlocked (WSTG-ATHN-03). Password change and reset are checked for whether one user can change or reset another user's password and whether a reset can be guessed or bypassed (WSTG-ATHN-09), and for whether reset skips a multi-factor sign-in that is turned on (ASVS v5.0.0-6.4.3).
Sessions are tested at the moments they should end. After sign-out, the old session is replayed to see whether the server still accepts it (WSTG-SESS-06; ASVS v5.0.0-7.4.1). After an administrator suspends an account, that account's open sessions should stop working as well (ASVS v5.0.0-7.4.2). Lockout and reset checks run only on test accounts, where the rules of engagement allow them.
Broken access control is first in the OWASP Top 10:2025, and inside a web application the place to start is records. Every route that names a record by id, whether in the path, the query string or the body, is sent again from a second account that does not own the record: reads, edits and, on test records only, deletes.
WSTG v4.2 describes the method as two users with identical privileges, both signed in, with the session identifier and parameters swapped between them on every request (WSTG-ATHZ-02). Its insecure direct object reference test adds that the best way to find these flaws is with at least two users who own different objects (WSTG-ATHZ-04).
ASVS 5.0 asks for data-specific access to be restricted to consumers with explicit permission to each data item (v5.0.0-8.2.2), and OWASP's API list names the same weakness API1:2023, broken object level authorisation.
Downloads get the same treatment, plus two more requests: one with no session and one after sign-out, because WSTG v4.2 checks whether each resource can be reached without authenticating and after logging out (WSTG-ATHZ-02). A generated PDF or an uploaded file can be served from a different route from the screen that shows it, and that route has to make the ownership check on its own.
Exports deserve a separate look, because an export can build its own query rather than reuse the one behind the screen.
Vertical tests take requests that only an administrator should be able to make and send them from a lower role's session and from no session at all (WSTG-ATHZ-02; ASVS v5.0.0-8.2.1). A route is not treated as protected or unprotected because of how its path looks.
OWASP's API guidance warns against judging an endpoint regular or administrative from its URL path alone, and asks whether a user can reach a sensitive action by changing the HTTP method, such as from GET to DELETE (API5:2023).
Two further checks matter in an application with paying customers. When an administrator changes a customer's plan or suspends them, the change is tested against a session the customer already has open, since ASVS 5.0 asks for changes to the values behind authorisation decisions to apply immediately (v5.0.0-8.3.2, a Level 3 requirement).
And sensitive administrator actions, such as issuing account credit, are checked for the safeguard WSTG v4.2 suggests for them: a maker-checker approval or multi-factor sign-in (WSTG-IDNT-01).
The screens are only one client of your API, and nothing obliges an attacker to use them. Each route on the list is called directly as well as through the interface, with fields the form never sends and values the browser would refuse. WSTG v4.2 describes this as forging requests that go around the front end and straight to back end processing (WSTG-BUSL-02).
Two questions about fields get their own test cases. Do responses carry properties the screen never shows, such as another person's email address, an internal flag or a password hash? And will the server accept a change to a property the user should not control, such as plan, role, credit balance or owner?
OWASP files both under API3:2023, broken object property level authorisation, and ASVS 5.0 asks for field-level access control and for limits on the fields each action accepts (v5.0.0-8.2.3, v5.0.0-15.3.3).
Routes that cost money each time they run, such as those that send email or text messages, are checked for a limit with a handful of requests to addresses the tester controls. OWASP notes that abuse of resources paid for per request, such as email and SMS, can increase operating costs (API4:2023). This is a check for a missing limit, not load testing, which no Clever Ops service includes.
API testing on this page means the API your own application calls, tested as part of the application. An API that partners or a mobile app call with their own keys or tokens is a separate surface with its own identities, so it is scoped as one, and the mobile app itself is not part of the service.
Business logic tests hold the application to the rules list. Values the browser would block are sent straight to the server, such as a line quantity below one, a percentage discount over its cap or an expiry date that has already passed (WSTG-BUSL-01; ASVS v5.0.0-2.2.2).
Steps are taken out of order or repeated, such as accepting a quote after it was withdrawn (WSTG-BUSL-06). Limits are pushed past their edge, such as a free plan's monthly allowance reached through the API rather than the screen (WSTG-BUSL-05; ASVS v5.0.0-2.3.2).
WSTG v4.2 says business logic flaws cannot be detected by a vulnerability scanner, which is the plainest reason a scan and a penetration test are different purchases.
Payment testing covers your application's side of the payment, not the payment provider's platform. Payment checks run against a staging copy linked to a non-live payment account wherever one is available, so no real card is charged. The tests ask three questions. Do the amount and the plan come from the server, or can the checkout request change them (WSTG-BUSL-02)?
Is an upgrade applied only once the provider confirms the payment, rather than when the browser reaches a success page? And does the route that receives the provider's notifications reject a message without a valid signature?
For a Stripe checkout, Stripe's own order fulfilment guide says fulfilment cannot rely on the checkout landing page alone, and its example webhook handler verifies the signature and rejects a request whose signature is invalid.
A payment test is not a PCI DSS assessment. Clever Ops is not a PCI Approved Scanning Vendor or a Qualified Security Assessor. External scans and assessments under PCI DSS are referred to providers that hold those roles.
Each upload point is tried with file types the business does not accept and with harmless files whose content does not match their extension, because checking the extension alone is the low-assurance form of the control (WSTG-BUSL-08; ASVS v5.0.0-5.2.2). Stored files are then requested directly to see whether any is run as server-side code (ASVS v5.0.0-5.3.1), and the processing that follows an upload, such as resizing an image or attaching a file to a generated PDF, is tried with harmless test files (WSTG-BUSL-09). Whether another account can download an upload is an access question, tested with the records above.
Anything one person writes and another person later views is checked for stored scripting, with a harmless marker in place of a script (WSTG-INPV-02). Paths that cross roles deserve the closest look: a customer name or note written by an ordinary account and shown in an administrator's screen gives that account a way to reach your staff. Search, filter and sort parameters are checked for SQL injection points (WSTG-INPV-05).
Worked scope: a twenty-route, two-role SaaS application
An illustration, not a client. The application, quotes.lab.test, is an invented quoting and invoicing product sold by subscription to sole traders, and its routes, accounts and every scope decision below were written for this page.
How to use itHide the instructions
It has two roles, subscriber and platform administrator, and administrators sign in with multi-factor authentication. Its twenty routes: R1 POST /api/auth/signup; R2 POST and DELETE /api/auth/session (sign in, sign out); R3 POST /api/auth/reset (request a reset email); R4 POST /api/auth/reset/confirm; R5 GET and POST /q/{token} (a subscriber's customer views and accepts a quote without signing in); R6 GET and PATCH /api/me; R7 POST /api/me/logo; R8 GET and POST /api/customers; R9 GET, PATCH and DELETE /api/customers/{id}; R10 GET and POST /api/quotes; R11 GET and PATCH /api/quotes/{id}; R12 POST /api/quotes/{id}/send; R13 GET /api/quotes/{id}/pdf; R14 POST /api/quotes/{id}/files; R15 GET /api/export/customers; R16 POST /api/billing/checkout; R17 POST /api/billing/webhook (called by the payment provider, Stripe); R18 GET /api/admin/subscribers; R19 PATCH /api/admin/subscribers/{id} (change plan, suspend); R20 POST /api/admin/subscribers/{id}/credit. The rules list: a free plan sends at most five quotes a calendar month; a quote can be accepted once, and not after it is withdrawn or expires; line quantities are at least one; discounts are capped at 30 percent; only administrators issue account credit. Test accounts: sub-a (free plan) and sub-b (paid plan), each with seeded customers and quotes, and admin-1 and admin-2, on a staging copy linked to a non-live payment account. Lines T1 to T23 are test cases and lines X1 to X12 are exclusions. To reuse it, replace the routes, rules and accounts with your own, keep every column, and give each line a scope decision.
35 rows
Line
Routes in the example
T1. Account enumeration
- Routes in the example
- R1, R2, R3
- What is tested, or why it is excluded
- Responses to sign-up, sign-in and reset requests compared for an email address with an account and one without: message text, status code and response time.
- Accounts, data and conditions
- sub-a, and an address the tester controls that has no account.
- OWASP reference
- WSTG-IDNT-04; ASVS v5.0.0-6.3.8
- Scope decision
- In scope
T2. Sign-in lockout
- Routes in the example
- R2
- What is tested, or why it is excluded
- Repeated wrong passwords on one account, then the right one, to see whether the account locks or slows further attempts, and how it unlocks.
- Accounts, data and conditions
- sub-b only, which the client has agreed may be locked. Guessing across many accounts is not attempted.
- OWASP reference
- WSTG-ATHN-03; ASVS v5.0.0-6.3.1
- Scope decision
- In scope
T3. Password reset
- Routes in the example
- R3, R4
- What is tested, or why it is excluded
- Whether one user can reset another user's password, whether a reset can be guessed or bypassed (WSTG-ATHN-09), and whether reset skips an administrator's multi-factor sign-in (6.4.3). Extra, no OWASP id: whether a reset link still works after it has been used or after a newer link is issued.
- Accounts, data and conditions
- sub-a, sub-b and admin-1, with reset emails going to inboxes the tester controls.
- OWASP reference
- WSTG-ATHN-09; ASVS v5.0.0-6.4.3
- Scope decision
- In scope
T4. Session end
- Routes in the example
- R2, R6, R19
- What is tested, or why it is excluded
- The session cookie from before sign-out replayed on GET /api/me (WSTG-SESS-06, 7.4.1), and again after admin-1 suspends the account through R19 (7.4.2).
- Accounts, data and conditions
- sub-a and admin-1. The suspension is reversed afterwards and listed in the report.
- OWASP reference
- WSTG-SESS-06; ASVS v5.0.0-7.4.1, v5.0.0-7.4.2
- Scope decision
- In scope
T5. Customers and quotes between subscribers
- Routes in the example
- R9, R11
- What is tested, or why it is excluded
- Each request naming a customer or quote owned by sub-a is sent from sub-b's session: read, edit and delete, then the same in reverse.
- Accounts, data and conditions
- sub-a and sub-b with seeded records. Deletes run only on records created for the test.
- OWASP reference
- WSTG-ATHZ-02 (horizontal), WSTG-ATHZ-04; ASVS v5.0.0-8.2.2; OWASP API1:2023
- Scope decision
- In scope
T6. Quote PDFs and attached files
- Routes in the example
- R13, and the file links R11 returns
- What is tested, or why it is excluded
- Another subscriber's PDF and attachments requested by id from sub-b's session, with no session, and after sign-out.
- Accounts, data and conditions
- sub-a's quotes with one attachment each, requested as sub-b and as a visitor.
- OWASP reference
- WSTG-ATHZ-02, WSTG-ATHZ-04; ASVS v5.0.0-8.2.2
- Scope decision
- In scope
T7. Customer export
- Routes in the example
- R15
- What is tested, or why it is excluded
- The export requested by sub-b with any account or filter parameter changed to point at sub-a, and the file checked for rows that belong to anyone else.
- Accounts, data and conditions
- sub-a and sub-b, each with customers whose names mark the owner.
- OWASP reference
- WSTG-ATHZ-04; ASVS v5.0.0-8.2.2
- Scope decision
- In scope
T8. Public quote link
- Routes in the example
- R5
- What is tested, or why it is excluded
- Extra, no OWASP id: whether a link token can be guessed or worked out from another quote's link, and whether the page shows anything beyond its own quote. Accepting the same quote twice (WSTG-BUSL-05, 2.3.2).
- Accounts, data and conditions
- Quotes created by sub-a and sub-b and sent to addresses the tester controls.
- OWASP reference
- WSTG-BUSL-05; ASVS v5.0.0-2.3.2 (repeat acceptance only)
- Scope decision
- In scope
T9. Administrator routes from a subscriber
- Routes in the example
- R18, R19, R20
- What is tested, or why it is excluded
- Each admin request, captured from admin-1's session, sent with sub-a's session and with no session, and again with the HTTP method changed where the route accepts more than one.
- Accounts, data and conditions
- admin-1 and sub-a. Any change that succeeds is reversed and listed for clean-up.
- OWASP reference
- WSTG-ATHZ-02 (vertical); ASVS v5.0.0-8.2.1; OWASP API5:2023
- Scope decision
- In scope
T10. Fields a subscriber should not control
- Routes in the example
- R1, R6, R11
- What is tested, or why it is excluded
- Sign-up, profile and quote updates sent with plan, role, credit balance, account status or owner fields added, and responses read for properties the screen never shows.
- Accounts, data and conditions
- A new sign-up made for the test, plus sub-a.
- OWASP reference
- WSTG-ATHZ-03, WSTG-BUSL-02; ASVS v5.0.0-8.2.3, v5.0.0-15.3.3; OWASP API3:2023
- Scope decision
- In scope
T11. Plan change during an open session
- Routes in the example
- R19, then R10 and R12
- What is tested, or why it is excluded
- admin-1 moves sub-b to the free plan while sub-b has a session open, and sub-b's next requests are checked against the free plan's limits.
- Accounts, data and conditions
- admin-1 and sub-b. The plan is restored afterwards.
- OWASP reference
- ASVS v5.0.0-8.3.2; WSTG v4.2 has no matching test, so the method is an extra with no WSTG id
- Scope decision
- In scope
T12. Issuing account credit
- Routes in the example
- R20
- What is tested, or why it is excluded
- Whether issuing credit asks for the safeguard WSTG suggests for sensitive admin functions, a maker-checker approval or multi-factor sign-in (WSTG-IDNT-01), and whether the server refuses a negative amount (WSTG-BUSL-01, 2.2.2).
- Accounts, data and conditions
- admin-1, admin-2 and sub-a. Credit issued in testing is reversed and listed.
- OWASP reference
- WSTG-IDNT-01, WSTG-BUSL-01; ASVS v5.0.0-2.2.2
- Scope decision
- In scope
T13. Routes beyond the list
- Routes in the example
- All twenty, compared with what is deployed
- What is tested, or why it is excluded
- The route list compared with the requests the front end makes and the routes that answer (WSTG-INFO-06, WSTG-INFO-07). An unlisted route, such as an older /api/v1 path, is reported and tested only once the client confirms it in writing.
- Accounts, data and conditions
- Every account above, plus no session.
- OWASP reference
- WSTG-INFO-06, WSTG-INFO-07; OWASP API9:2023
- Scope decision
- In scope. Found routes added only by written agreement
T14. Routes that send messages
- Routes in the example
- R3, R12
- What is tested, or why it is excluded
- A handful of reset and quote emails sent within the agreed request rate, to see whether any limit applies per account or per recipient. Volume testing is not attempted.
- Accounts, data and conditions
- sub-a, sending only to addresses the tester controls.
- OWASP reference
- OWASP API4:2023
- Scope decision
- In scope
T15. Values the server should refuse
- Routes in the example
- R11
- What is tested, or why it is excluded
- Quote lines saved through the API with a quantity below one, a discount above the 30 percent cap and an expiry date already passed, and the total compared with the server's own calculation.
- Accounts, data and conditions
- sub-a and sub-b, on their own quotes.
- OWASP reference
- WSTG-BUSL-01; ASVS v5.0.0-2.2.2
- Scope decision
- In scope
T16. Free plan allowance
- Routes in the example
- R10, R12
- What is tested, or why it is excluded
- sub-a sends a sixth quote in one month through the API rather than the screen, including two sends fired at nearly the same moment.
- Accounts, data and conditions
- sub-a on the free plan, with five quotes already sent that month.
- OWASP reference
- WSTG-BUSL-05; ASVS v5.0.0-2.3.2. The two sends at the same moment: ASVS v5.0.0-15.4.2
- Scope decision
- In scope
T17. Quote acceptance out of order
- Routes in the example
- R5, R11
- What is tested, or why it is excluded
- A withdrawn quote and an expired quote accepted by replaying their accept requests, and an accepted quote edited by its subscriber afterwards.
- Accounts, data and conditions
- Quotes from sub-a and sub-b in each state.
- OWASP reference
- WSTG-BUSL-06
- Scope decision
- In scope
T18. Checkout amount and plan
- Routes in the example
- R16
- What is tested, or why it is excluded
- The checkout request sent with the plan id or price changed, to see whether the amount charged comes from the server.
- Accounts, data and conditions
- sub-a, on staging linked to a non-live payment account, so no real card is charged.
- OWASP reference
- WSTG-BUSL-02
- Scope decision
- In scope
T19. Payment confirmation
- Routes in the example
- R16, R17
- What is tested, or why it is excluded
- Extra, no OWASP id: whether the upgrade happens only after the provider's webhook confirms payment, not when the browser reaches the success page, and whether a webhook message with no valid signature is rejected. Stripe's order fulfilment guide describes both controls for a Stripe checkout.
- Accounts, data and conditions
- sub-a, on the non-live payment account. Webhook messages go only to the staging copy.
- OWASP reference
- None: extra checks, measured against the payment provider's own guidance
- Scope decision
- In scope. The hosted checkout page is excluded (X6)
T20. File type and content
- Routes in the example
- R7, R14
- What is tested, or why it is excluded
- Logo and attachment uploads tried with file types the business does not accept and with harmless files whose content does not match their extension.
- Accounts, data and conditions
- sub-a. Test files only, none of them harmful.
- OWASP reference
- WSTG-BUSL-08; ASVS v5.0.0-5.2.2
- Scope decision
- In scope
T21. What happens to stored files
- Routes in the example
- R7, R14, R13
- What is tested, or why it is excluded
- Stored uploads requested directly to see whether any is run as server-side code (5.3.1), and the steps after upload, resizing the logo and adding attachments to the quote PDF, tried with harmless test files (WSTG-BUSL-09).
- Accounts, data and conditions
- sub-a. Uploaded test files are deleted or listed for clean-up.
- OWASP reference
- WSTG-BUSL-09; ASVS v5.0.0-5.3.1
- Scope decision
- In scope
T22. Stored input across roles
- Routes in the example
- R8, R11, then R5 and R18
- What is tested, or why it is excluded
- A harmless marker saved in customer names and quote notes, then viewed where other people see it: the customer's public quote page and the administrator's subscriber list.
- Accounts, data and conditions
- sub-a writes; a visitor and admin-1 view.
- OWASP reference
- WSTG-INPV-02
- Scope decision
- In scope
T23. Search, filter and sort parameters
- Routes in the example
- R8, R10, R18
- What is tested, or why it is excluded
- Each search, filter and sort parameter checked for SQL injection points.
- Accounts, data and conditions
- sub-a and admin-1, on seeded records.
- OWASP reference
- WSTG-INPV-05
- Scope decision
- In scope
X1. Team accounts
- Routes in the example
- None in the example
- What is tested, or why it is excluded
- Several users with their own roles inside one subscriber account would add columns to the permission matrix and new pairs of roles to test.
- Accounts, data and conditions
- Would need two accounts for each added role.
- OWASP reference
- None: an exclusion
- Scope decision
- Excluded. Adds effort, so it is scoped separately
X2. Records shared between organisations
- Routes in the example
- None in the example
- What is tested, or why it is excluded
- Subscribers sharing customers or quotes with each other would need cross-tenant checks in both directions for every shared record (ASVS v5.0.0-8.4.1).
- Accounts, data and conditions
- Would need accounts in at least two organisations that share records.
- OWASP reference
- ASVS v5.0.0-8.4.1, for the check it would add
- Scope decision
- Excluded. Adds effort, so it is scoped separately
X3. Single sign-on
- Routes in the example
- None in the example
- What is tested, or why it is excluded
- Signing in through a subscriber's own identity provider adds the sign-in exchange, account linking and the provider settings to test.
- Accounts, data and conditions
- Would need a test identity provider the client controls.
- OWASP reference
- None: an exclusion
- Scope decision
- Excluded. Adds effort, so it is scoped separately
X4. Mobile and partner clients
- Routes in the example
- The same API, called with other keys
- What is tested, or why it is excluded
- The API routes the web application uses stay in T1 to T23. Keys and tokens issued to a mobile app or partners are a separate surface, and the mobile app itself is not part of the service.
- Accounts, data and conditions
- Would need the partner or mobile keys, and the client's list of what each may call.
- OWASP reference
- None: an exclusion
- Scope decision
- Excluded. Partner and mobile API access scoped separately; the mobile app itself is not offered
X5. AI assistant in the product
- Routes in the example
- None in the example
- What is tested, or why it is excluded
- An assistant that drafts quotes or answers customers needs prompt injection, retrieved content and its actions tested as their own surface.
- Accounts, data and conditions
- Would need the assistant's tools, data sources and permissions listed.
- OWASP reference
- None: an exclusion
- Scope decision
- Excluded. Scoped as an AI chatbot and agent security test
X6. Hosted checkout and payment platform
- Routes in the example
- The provider's pages, outside the twenty routes
- What is tested, or why it is excluded
- The checkout page, card handling and the payment provider's dashboard belong to the provider. The application's side of the payment is tested in T18 and T19.
- Accounts, data and conditions
- No access to the provider's systems is sought.
- OWASP reference
- None: an exclusion
- Scope decision
- Out of scope. Owned by the payment provider
X7. PCI DSS scans and assessments
- Routes in the example
- Not applicable
- What is tested, or why it is excluded
- External scans and assessments under PCI DSS are a different service. Clever Ops is not a PCI Approved Scanning Vendor or a Qualified Security Assessor.
- Accounts, data and conditions
- Not applicable.
- OWASP reference
- None: an exclusion
- Scope decision
- Referred to a provider that holds those roles
X8. Hosting platform, servers and office network
- Routes in the example
- Not web routes
- What is tested, or why it is excluded
- The hosting provider's platform belongs to the host. Internal networks, servers and staff devices are not part of any Clever Ops service.
- Accounts, data and conditions
- Not applicable.
- OWASP reference
- None: an exclusion
- Scope decision
- Host platform out of scope. Networks and devices referred to a network specialist
X9. Load and denial-of-service testing
- Routes in the example
- Any route
- What is tested, or why it is excluded
- Testing how much traffic the application can take is outside every Clever Ops service.
- Accounts, data and conditions
- Would need the host's written approval.
- OWASP reference
- None: an exclusion
- Scope decision
- Referred to a load-testing specialist
X10. A tender that requires CREST
- Routes in the example
- Not applicable
- What is tested, or why it is excluded
- Where a tender, panel or contract requires a CREST-accredited provider, the work goes elsewhere. Clever Ops is not CREST accredited, so we refer that work to a provider that is.
- Accounts, data and conditions
- Not applicable.
- OWASP reference
- None: an exclusion
- Scope decision
- Referred to a CREST-accredited provider
X11. Full secure code review
- Routes in the example
- The whole repository
- What is tested, or why it is excluded
- Read-only code access to named areas, such as the R17 webhook handler, can be added to this scope. A line-by-line review of the repository is a separate piece of work.
- Accounts, data and conditions
- Would need repository access and the client's list of priority areas.
- OWASP reference
- None: an exclusion
- Scope decision
- Excluded unless agreed. Scoped separately
X12. Features released after the scope is agreed
- Routes in the example
- Any new route
- What is tested, or why it is excluded
- A route or feature that appears after the scope is signed was not part of the test, and the retest does not cover it.
- Accounts, data and conditions
- Would need the route added to the list and the matrix.
- OWASP reference
- None: an exclusion
- Scope decision
- Excluded. Scoped separately as new work
Test accounts, data and conduct on a working application
The scope asks for two test accounts in each role, each with sample records, so access can be tested in both directions between people in the same role.
If your application serves separate customer organisations, it asks for accounts in two of them as well. A staging copy that matches production is preferred. When production is the target, the authorisation letter names it and records the risks you accept.
Testing starts only once the owner of each target has signed an authorisation and your hosting provider's testing policy has been checked. The conduct rules are the same on every engagement:
- Destructive actions are off by default for real people's accounts and records that existed before the test.
- Reset, lockout and deletion checks run only on test accounts and test records, where the rules of engagement allow them.
- Nothing is sent by email or text message to a real person. Routes that send messages are tested with addresses the tester controls.
- Requests stay under an agreed rate, and testing on a target pauses if a firewall starts blocking or your contact asks for a pause.
- Test data is tracked as it is created, then removed or listed in the report for your team to remove.
- Evidence has session tokens, passwords, keys and personal details redacted, is stored encrypted in Australia, and is deleted on the schedule set in the data handling terms.
How much access the tester gets
Testing is authenticated by default, because most of the areas above sit behind the login. Read-only code access to named areas, such as the permission checks or the payment handler, can be added to confirm how a control is enforced, and the report marks which findings were confirmed on the running build. The comparison of black box, grey box and white box testing explains what each level of access changes.
Features that change the scope, and work that goes elsewhere
The exclusions in the worked scope fall into five groups.
When you read a quote, check which group each exclusion belongs to: some are effort that was left out, some are another business's system, and some are work for a different kind of provider.
- Scoped separately, because they add effort: teams with their own roles inside one customer account, customer organisations that share records with each other, single sign-on through customers' own identity providers, partner or mobile clients calling the API with their own keys, and features released after the scope is agreed. Very logic-heavy products, such as marketplaces, lending platforms and multi-tenant billing, are scoped individually or referred to a larger firm.
- Scoped as a different piece of work: a full secure code review of the repository, and an AI assistant built into the product, which is tested under the AI chatbot and agent security test because prompt injection and the actions the assistant can take need their own test cases.
- Out of scope because another business owns it: the payment provider's hosted checkout and platform, your hosting provider's platform, and third-party services on your subdomains, unless their owner agrees.
- Referred to a specialist: internal networks, servers and staff devices go to a network specialist, and load or denial-of-service testing goes to a load-testing specialist with your host's written approval. Where a tender, panel or contract requires CREST, the work goes elsewhere too. Clever Ops is not CREST accredited, so we refer that work to a provider that is.
- Outside this service altogether: live Shopify storefronts and Laravel applications.
If Clever Ops built or hosts the application
The same test cases then run as a security review rather than an independent penetration test, and the report states who built the application and who hosts it. When the person relying on the report asks for independence, choose a provider that had no part in building or running it.
What the report and retest give you
The report follows the scope you agreed, so you can hold one against the other.
The penetration test report example shows the format, using a lab application rather than a client's.
- The scope as tested: the build, the dates, the route list, the roles and accounts used, and each exclusion.
- Findings with redacted evidence, the accounts that produced it, business impact, a fix and the criteria the retest will use. Each rated finding carries a CVSS v3.1 score and a business-risk rating.
- Negative results for the routes and accounts tried, so a later question, such as whether one customer could read another customer's quotes, can be answered from the record.
- Test cases that could not run, with the reason, such as an integration that was down on staging or an account that was never provided.
- A list of any test data left behind for your team to remove.
Fixing the findings and the one retest
You choose who fixes the findings: your own developers, your agency, or a Clever Ops security fix sprint under its own statement of work. One consolidated retest is included. You request it within the window set in your statement of work, once the fixes are deployed to an identified build of the same application, and it covers the reported findings within the agreed scope.
Each finding then ends as fixed, partially fixed, still reproducible or not retested, and a finding left out of the retest is reported as not retested rather than as passed.
Frequently asked questions
Not always. Some providers use web application security assessment for a scan with a short review, and others for a full signed-in test. Ask whether the work signs in as every role, calls the API directly and tests your business rules. If it does, it is a penetration test, whatever it is called. If it stays outside the login, it is closer to a website security check.
Yes. The API your application's own front end calls is tested directly as part of the application, including routes that no screen uses. Older versions that still answer are reported and added to the test once you confirm them in writing. An API that partners or a mobile app call with their own keys or tokens is scoped as its own surface, and the mobile app itself is not part of the service.
Two in each role in scope, each with some sample records, so access between people in the same role can be tested in both directions. If your application serves separate customer organisations, add accounts in two of them. For the worked scope on this page, that means two subscribers and two administrators.
The method stays the same and the scope grows with the application: more routes, more pairs of roles, more records to cross-check. Larger and multi-tenant applications are scoped individually after a longer call. Very logic-heavy products, such as marketplaces or lending platforms, may be referred to a larger firm.
Yes, under the conduct rules on this page, although a staging copy that matches production is preferred. On production, destructive actions are off by default for real people's accounts and records that existed before the test, requests stay under an agreed rate, and test data is removed or listed for your team to remove.
No. A penetration test exercises the running application, and a code review reads its source. WSTG v4.2 notes that dynamic testing has difficulty reaching an entire code base, and that source review is particularly suited to finding flawed business logic and access control problems. Read-only code access to named areas can be added to a web application test, while a full review of the repository is a separate piece of work.
Australian businesses that run a custom web application with customer or staff logins, such as a portal, a booking or quoting product, or a software product sold by subscription. Clever Ops is based in Gippsland, Victoria, and works with businesses across Australia. If your site has no customer logins, start with the website security check. A WordPress site is better served by the WordPress security audit, and an app made with an AI builder by the vibe-coded app security check.
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.