Penetration Testing · 9 min read

Business logic penetration testing: verify your portal’s rules

Test whether a customer portal preserves pricing, approvals, and reservation limits when actions are changed, skipped, or repeated. Define the rules first, then inspect evidence from the application’s actual state.

THE BRIEF

What you’ll take away

  • Turn business expectations into explicit rules for prices, roles, quote versions, approvals, and reservation state.
  • Test allowed journeys alongside changed, skipped, repeated, and bounded concurrent requests, using an authorized environment and synthetic records.
  • Verify reservation and inventory outcomes behind the response, then report what passed, failed, or remains untested.
Explore the guide
In this guide

Start with the outcome the business must protect

A fictional equipment-rental company is preparing a customer portal. A contractor selects equipment and rental dates, receives a quote, requests a discount where needed, and confirms a reservation after approval. The company needs those steps to preserve its commercial rules: customers receive the correct price, only authorized staff approve exceptions, and equipment is committed once for each valid reservation. The security question is whether those rules still hold when someone uses the application in an unexpected sequence.

Business logic penetration testing examines that relationship between the intended process and the application’s behavior. OWASP’s Web Security Testing Guide describes workflow testing as developing application-specific misuse cases from requirements and use cases, including attempts to complete a process without the required steps. [1] For the rental company, that means reviewing more than whether a user can open a page. The important outcome is whether a reservation can exist without the price, approval, and inventory conditions the business requires.

Choose a small set of valuable journeys and identify their owners. Product may define the customer experience, operations owns equipment availability, finance defines pricing rules, and engineering implements the transitions. Bring them together before testing. If two owners disagree about when a discount expires or whether a changed quote needs approval, resolve the policy first. A tester needs a clear expected outcome to distinguish a security failure from an undocumented but intentional business decision.

Draw the workflow as states and permitted changes

For this fictional portal, use four working states: quote prepared, discount approval pending, current quote approved, and reservation confirmed. The illustrated path concerns a quote whose discount requires approval; an ordinary quote may have a different permitted path. Define who can make each transition and the conditions that must be true. A customer can request an exception, an authorized rental employee can approve it, and the customer can confirm only when the current quote meets the required conditions.

Then describe changes that move the quote backward. In this example, changing equipment, rental dates, quantity, or the requested discount invalidates the existing approval and returns the quote to pricing and review. These are proposed business rules for the example, not universal requirements for every rental service. Decide which edits are material in your own application and how the application records that distinction. Keep customer ownership and staff approval authority explicit at every transition.

Include incomplete and abandoned work. Does a quote expire? Does preparing one reserve equipment, or does inventory change only at confirmation? What happens when a customer cancels or a background task fails? Agree on the expected state and any cleanup before turning these into test cases. This makes the workflow useful to both product and engineering: it describes the records and commitments the business expects, including when an action does not finish normally.

01 / FOLLOW THE BUSINESS STATE

Approval belongs to a specific quote.

  1. PrepareQuote v2

    Discount exception awaits staff review.

  2. AuthorizeApproved v2

    Authorized staff approves these exact terms.

  3. ConfirmReservation

    Customer confirms the currently approved version.

Rental dates change → quote v3

Return to repricing and review. Obtain approval for the revised terms before confirmation.

Synthetic quote Q-104 in a fictional rental portal. These are proposed rules: v3 must not inherit v2’s approval.

Make price and approval authority explicit

OWASP’s Business Logic Security guidance recommends deriving security-relevant values, including prices and permissions, from trusted server-side information instead of accepting them as authoritative client input. [2] For the rental portal, identify the source of the rate, account terms, rental duration, quantity, and approved discount. The browser can display an estimate and collect selections, but the final commercial decision needs to use the values the service is authorized to apply.

Write acceptance rules in business language. A customer cannot choose an arbitrary final total. A discount outside their allowed terms needs approval from an authorized staff role. Approval belongs to the relevant customer and quote version. Confirming a reservation must check the current price and approval conditions. The person reviewing the rule should be able to explain which source establishes each value and which component enforces it before equipment is committed.

Use a normal transaction as the comparison case. Capture the expected price for a synthetic account, equipment selection, dates, and quantity under a fixed test rate schedule. Then vary one relevant condition at a time within the authorized test plan. Observe whether the service recalculates or rejects the request as intended. A difference from the displayed price is not automatically a vulnerability; establish whether the result violates the agreed pricing policy and whether it reaches a meaningful business record.

02 / TURN RULES INTO CHECKS

Write the expected business effect.

Price calculation
Test case

Alter the client-supplied total.

Expected effect

Recalculate from validated terms.

Approval authority
Test case

Customer attempts staff approval.

Expected effect

Reject; retain the pending state.

Current-version approval
Test case

Confirm v3 with v2 approval.

Expected effect

No reservation or inventory commitment.

Retry behavior
Test case

Repeat the same valid confirmation.

Expected effect

One reservation; one inventory commitment.

Proposed rental-policy checks, not test results. A retry may return the original success; assess whether it creates a duplicate business effect.

Check what an approval actually authorizes

An approval is useful only if the executed action matches what was approved. OWASP’s Transaction Authorization guidance calls for checking authorization at execution and addresses changing transaction data during the authorization process. [3] Apply that principle to the rental workflow: the final confirmation needs a valid approval for the current material terms. A flag saying that somebody approved the quote at an earlier point cannot answer whether the approved terms still match.

Use synthetic quote Q-104 as a planned test case. Version 2 has staff approval for its original rental dates. Version 3 changes those dates, which this example treats as a material edit. The expected result is that confirmation remains unavailable until the new version is reviewed and approved. Test the underlying confirmation interface as well as the visible journey, because a disabled button alone does not establish that the service enforces the rule.

Include the identity dimension. The customer should not approve their own discount exception, another customer should not reuse the approval, and staff should act only within their assigned authority. These checks overlap with access control: a legitimate role may still be unauthorized for this particular quote, amount, or state. Keep the account, quote version, approving role, and timing visible in the test record so a later reviewer can understand why the action should have been allowed or denied.

Define what retries and competing requests should do

Customers retry when a page stalls, and integrations can repeat work after losing a response. OWASP’s function-use-limit scenario recommends identifying actions with a permitted number of executions and testing whether that limit is enforced. [4] For the fictional portal, define confirmation as one business operation: retrying the same confirmation should not create another reservation or consume inventory a second time. A second, deliberately separate booking is a different case and needs its own availability checks.

Concurrent requests add another question: can two operations both proceed based on the same remaining equipment? OWASP’s Business Logic Security guidance discusses protecting check-and-update operations and coordinating repeated requests, including their identity and parameters. [2] Agree on a bounded concurrency test for a synthetic inventory item. The expected outcome is that confirmed commitments respect available capacity. Do not treat a browser throttle or one sequential denial as evidence about simultaneous backend operations.

Define the retry window and recovery behavior with engineering. If the application cannot tell whether an external inventory action succeeded, it should have an agreed way to resolve that uncertainty before repeating the effect. Test what the customer sees and what the inventory system records. Keep request volume and side effects within the authorized limits. These checks investigate consistency of the business operation; they are not permission to stress a live reservation service or exhaust real equipment availability.

Prepare a test environment that can reveal the result

Use representative test accounts for customers and the relevant staff roles, synthetic equipment, a known rate schedule, and a small inventory pool reserved for testing. The environment needs enough production-like behavior to exercise the rules being assessed. Record meaningful differences, such as a simulated inventory integration or disabled notification worker. A result from a simplified environment can still be useful, but its limitations must remain attached to the conclusion.

Agree on authorized applications, interfaces, test windows, permitted side effects, stop conditions, and cleanup. Identify any third-party services and confirm the testing arrangements that apply to them. In the rental example, the team should know whether a test can trigger a supplier message, create a dispatch task, or alter an inventory feed. Arrange access to relevant records or operator-assisted verification so testers can observe the business outcome without needing unrestricted production access.

Also make the starting state reproducible. Record the build, configuration, account permissions, quote version, and relevant background-job status. If a test changes the synthetic inventory pool, restore or account for that change before the next case. Otherwise a later request may fail because of test contamination rather than because the control worked. A well-prepared environment helps reviewers separate a reproducible business-rule issue from an unrelated outage or inconsistent test setup.

Related: settle the scope before a penetration test

Connect the request to the business record

An evidence record should connect the rule, starting state, action, response, and resulting application state. For planned case BL-04, record customer C-17, quote Q-104 version 3, and the earlier approval for version 2. The action is a confirmation attempt after the rental-date change. The expected result is rejection, with no new reservation or inventory commitment. Mark the result Not run until an authorized test has actually produced evidence. These identifiers describe a fictional test plan, not an SPS assessment outcome.

During execution, retain the relevant request and response, quote and approval versions, reservation records, and inventory state before and after the action. Correlate the observations with timestamps or identifiers and protect sensitive material in the evidence package. If work is asynchronous, define an observation window and inspect the relevant completion or failure records. A quick error response does not establish that a queued action stopped, and an immediate success message does not prove that inventory was durably committed.

State the conclusion only as strongly as those observations allow. An unauthorized reservation record can support a finding about reservation enforcement; it does not by itself prove that equipment was dispatched, money was lost, or a real customer was affected. If the inventory integration could not be observed, report that gap. For a denied case, retain the absence of the prohibited state change within the checked scope and observation window alongside the application response.

03 / CHECK THE RESULTING STATE

Connect the request to the business records.

Expected rule / Q-104Confirmation of v3 requires renewed approval.
Request evidencePreserve the versions.

Record the request identifier, quote v3, referenced approval v2, and response.

Resulting stateInspect the records.

Check reservation creation, inventory commitments, and approval state.

Not tested

After an authorized run, compare observed records with the expected state and a valid confirmation control case.

Evidence plan only. A response code alone cannot establish the outcome. Report the demonstrated result and any remaining uncertainty.

Turn a failed rule into an owned engineering decision

A useful finding explains the business rule and the precise conditions under which it failed. Identify the role, quote state, affected interface, resulting record, and evidence limits. Describe impact through a supported scenario, such as reserving equipment under terms that lack the required approval. Keep possible downstream consequences separate from demonstrated effects. Product and operations can then help engineering judge priority using the affected workflow, exposure, and realistic consequence.

Assign the correction to the component that owns the decision. Depending on the evidence, the change might involve price derivation, quote-version binding, a final approval check, or coordination of inventory updates. A front-end message may improve the experience while leaving the underlying rule unenforced. Ask the implementation owner to explain where the rule will be checked and how competing or repeated requests will be handled. The test report should make that discussion specific without prescribing an unverified patch.

Retest the affected misuse case and an allowed journey after the change. For Q-104, that means checking rejection while approval is stale and successful confirmation after the current version receives valid approval. Preserve the tested build and result, including any conditions that remain unresolved. Add suitable regression coverage to the application’s own test process. A scoped retest verifies the affected behavior under stated conditions; it does not establish that every pricing or reservation path is secure.

Related: turn penetration-test findings into verified fixes

Scope the assessment around real customer journeys

Before commissioning the work, bring a walkthrough of the portal, the important roles, interface documentation, and the rules that protect value or availability. Include exceptions that ordinary demonstrations rarely show: revised quotes, canceled work, expired approvals, interrupted confirmations, and repeated requests. Prioritize the journeys that matter most to the business. Agree on which combinations will be examined and how untested paths will be reported rather than assuming that every possible sequence can be exhausted.

Source Point Security’s Web Application Penetration Testing includes manual business-logic testing across agreed user journeys and provisioned roles, alongside access control, authentication, input handling, and other application security checks. Findings include reproduction evidence, severity, and developer-focused fixes, with an executive summary and retesting of remediated findings. The application’s own interfaces are examined as part of those flows. A standalone API serving partners or other consumers may warrant a separately scoped API Penetration Testing engagement.

For the rental portal, a useful release discussion would identify the tested rules, the evidence supporting each result, open findings, and remaining coverage limits. Leadership can then decide whether to proceed, restrict a feature, or complete further work. The assessment supports that decision with evidence about actual behavior. It should leave the team able to explain why the approved pricing and reservation process can be relied on within the tested boundary, and exactly what still needs attention.

Explore Web Application Penetration TestingExplore API Penetration Testing

See what useful reporting looks like. Our illustrative sample pentest report shows scope, evidence, risk, remediation, and verification using a fictional environment.

YOUR NEXT STEP

Test the rules behind your customer portal

Bring your important workflows, user roles, and acceptance rules. We can scope a web application penetration test around the business outcomes you need to verify.

Discuss Web Application Penetration Testing

References & further reading

  1. OWASP WSTG v4.2, Testing for the Circumvention of Work Flows
  2. OWASP, Business Logic Security Cheat Sheet
  3. OWASP, Transaction Authorization Cheat Sheet
  4. OWASP WSTG v4.2, Test Number of Times a Function Can Be Used Limits

Share this guide on LinkedIn