A vulnerability scanner reports no critical findings, so management calls the application secure. The report cannot tell whether a clerk can approve their own payment, whether one customer can download another's records or whether a former employee still has a valid session. Those are business-rule and authorization questions that require specific tests. A proper security review combines technical scanning with evidence about how the product is meant to protect people and data.
The scanner may have done useful work. It cannot establish that employees see only the customer records they are permitted to see, that financial approvals require the right person, or that the business has a safe procedure for recovering from a compromised administrator account.
Those questions depend on how the product is intended to work. A security test that does not understand the rules of the application may never ask them.
Begin with the authorization story
Consider a scheduling platform used by several businesses. Every business has an administrator, employees and customer records. A normal employee may view appointments assigned to them but should not export another business's entire client list.
A scanner might confirm that the export endpoint returns a valid response when an authorized session is present. That says little about whether it accepts a request from the wrong account. An effective review must attempt to cross the boundaries that the application claims to enforce.
The OWASP Application Security Verification Standard provides testable security requirements for web applications and services. Version 5.0.0 was released in May 2025. The standard is more suitable for detailed technical verification than treating a short list of common vulnerabilities as a complete assurance program.
Start by documenting roles, ownership and permissions in ordinary language. Who may create an account? Who may modify another user's data? Which actions require a second approval? Can a former employee access anything after their account is removed?
Then translate those promises into tests. A user from Organization A attempts to access Organization B's records, including through direct API calls rather than the user interface. An employee tries a manager-only action. A session is revoked and tested again. The expected result should be recorded before the test is run.
Security exists in the boundary, not in whether a button is hidden.
A scanner sees requests, not business consequences
Automated scanning can identify insecure configurations, some forms of injection and other recognizable weaknesses. It is especially useful for detecting regressions and providing repeatable coverage across a large codebase.
But imagine a payment approval endpoint that permits a clerk to approve their own expense. The HTTP request may be perfectly formatted, encrypted in transit and free of injection. The defect is a missing business rule.
The OWASP guidance on establishing a modern security program distinguishes broad awareness materials from the comprehensive, verifiable requirements in ASVS. It explicitly discourages claiming full coverage of the OWASP Top 10 through tools alone.
A complete review therefore includes design discussion, code inspection, access-control testing and observation of real workflows. A scanner becomes one source of evidence inside that process.
It is reasonable for a small team to begin with a limited scope. The mistake is presenting that limited scope as proof the entire application is secure.
Data ownership must be checked at every layer
A user interface can show only permitted records while an API returns every record unless a client supplies a filter. This is a common kind of authorization design error: the browser enforces a restriction that the server should own.
Imagine a request to fetch invoices using a numeric invoice ID. If the server checks only that the requester is logged in, a user may be able to request another customer's invoice by changing the identifier. The application needs to verify the user's relationship to the object being requested.
That verification should be consistent for reading, editing, exporting and deleting. A system may secure the ordinary detail view while overlooking a background download route or bulk export.
For multi-tenant software, tenant boundaries deserve especially explicit tests. Data isolation can involve database queries, object storage paths, caches, search indexes, scheduled jobs and analytics. A correct filter in one API is not evidence that every component uses the same protection.
Test negative cases intentionally. “Authorized user can view their invoice” is only half the requirement. “Unrelated authorized user cannot view it” reveals the boundary.
Defaults, secrets and recovery deserve attention
A product can have strong application code yet be deployed with dangerous configuration. Administrative credentials may be shared among employees. API keys can appear in client-side bundles or logs. Old service accounts may retain access after integrations are removed.
A review should identify where secrets are stored, who can rotate them, how permissions are scoped and what happens if one is exposed. The goal is not to display secret values in an audit report. It is to verify that custody and rotation processes exist and work.
Backups and recovery also belong in security review because a compromised environment may require rebuilding a clean system. A backup that cannot be restored cannot support the incident-response plan.
The environment's change process matters as well. Who can deploy code? Does the deployment record show which version went live? Can an authorized operator roll back safely? Are production credentials available to every developer or only to the services that need them?
None of those questions can be answered from a successful vulnerability scan alone.
Logs must help without becoming another risk
Security monitoring needs evidence of significant events: failed logins, privilege changes, rejected access attempts, sensitive exports and unusual administrative actions.
The OWASP Logging Cheat Sheet recommends defining events clearly, testing logging during security verification and protecting logs from injection and unauthorized access. The point is to make evidence trustworthy and useful.
Logging everything indiscriminately is not a good answer. Full HTTP request bodies can contain passwords, personal information and payment details. High-volume debug output can fill disks or increase monitoring costs. A secure design records relevant events while minimizing exposure and controlling retention.
Logs must also survive the kinds of incidents they are intended to explain. If an attacker who compromises an application can quietly erase the entire audit trail, an investigator may have little reliable evidence afterward.
The OWASP Top 10:2025 discussion of logging and alerting failures emphasizes events, integrity controls and alerting. A log nobody reads is less useful than a defined set of events tied to an incident-response process.
A review should test that the expected record appears when a sensitive action occurs and that a person or automated control can react when necessary.
Test the unhappy path, not only the ordinary path
Many weaknesses appear when expected information is missing, duplicated or inconsistent. What if a payment request is submitted twice? What if a file upload fails halfway through? What if a user changes organizations while a background task still runs under an old identity?
A review should use realistic error cases and verify that sensitive actions fail safely. For financial workflows, a retry must not execute a transfer twice. For account-management workflows, the system must not grant broad privileges merely because a role lookup failed.
Look for design assumptions written nowhere except in a developer's memory. Those are difficult to test and easy to lose when staff change. Convert them into documented requirements with pass/fail evidence.
The correct scope depends on the application. A static public magazine has a different threat model from a payment platform. Its risks may involve unauthorized publishing, compromised administrative access or malicious advertising scripts rather than inter-tenant invoices.
A good reviewer narrows the scope based on real assets and threats. They do not apply the same checklist indiscriminately and call the result comprehensive.
What a credible security report contains
A useful report identifies the tested application version, environment and dates. It states what was in scope, what was excluded, which methods were used and what credentials were available to the testers.
Findings should include the affected functionality, evidence of the weakness, likely impact, severity reasoning and remediation guidance. Sensitive details should be shared securely with the responsible team. A report does not need to publish working exploitation instructions to demonstrate that a boundary failed.
A remediation should be retested. “Developer says fixed” and “reviewer observed the corrected boundary” are different statements. Remaining risks and assumptions should be documented rather than hidden to make a score look clean.
If a team uses ASVS, it can map tests to relevant verification requirements and identify gaps. That makes the conclusion more precise: certain controls were verified under defined conditions. It does not mean the application is permanently immune to attack.
A realistic starting scope for a small team
Select one critical workflow and trace it from browser to database. Map the identities involved, the permissions required, the data stored and the logs produced. Identify the actions that would be most harmful if a user or service were allowed to perform them incorrectly.
Create test cases for valid access, unauthorized access, revoked access, malformed input and repeated requests. Run the relevant automated checks. Then have someone other than the feature's original developer review the authorization logic and the evidence.
Document the result, correct the defects and run the tests again. Add repeatable checks to the development pipeline where possible. Expand to the next workflow using the lessons learned.
This does not replace a specialist penetration test for a high-risk service. It provides a more honest and useful basis for deciding when specialist review is needed.
The difference between a scan and a security review is the question being asked. A scan asks whether known patterns are visible. A review asks whether the system protects the specific activities and data people depend on.
References and background
Our explainers link to source documents so readers can explore the evidence and interpret it in context. Websites and policies may change after publication.
- OWASP, *Application Security Verification Standard*asvs.owasp.org
- OWASP, *Establishing a Modern Application Security Program*top10.owasp.org
- OWASP, *Logging Cheat Sheet*cheatsheetseries.owasp.org
- OWASP, *Security Logging and Alerting Failures*top10.owasp.org