An e-commerce site is less a shop window than a transaction system: catalogue, cart, coupons, payment, shipping and returns meet in one application. That makes it risky to shrink security testing to a list that claims “these ten flaws exist, identically, in every store”. This article makes no such claim. It takes the OWASP Top 10:2025 categories and maps each one to concrete examples often seen in storefronts, account areas and payment flows. The aim is to make a test plan clearer, not to invent percentages or a homemade ranking of “the ten most dangerous flaws”.

OWASP notes two new categories and one consolidation in 2025; server-side request forgery (SSRF) now sits under broken access control. The rates in the official introduction come from contributed testing data and do not mean “this is how often it happens in your shop”. Even so, the order of the categories is a sound way to structure an e-commerce test plan.

1. Broken access control

A01:2025 Broken Access Control remains first. In the contributed data OWASP reports, at least one of the CWEs in this category was seen in 3.73% of applications tested, on average. In e-commerce the practical form is often missing object-level authorisation: viewing another customer’s order, invoice or address by changing an identifier; writing an admin coupon from an ordinary session; changing stock or price through an unauthenticated API call.

“I logged in and I cannot see the menu” is not enough. For every object, the question “does this record really belong to me?” should be asked in the web interface and in the API layer.

2. Security misconfiguration

A02:2025 Security Misconfiguration rose from fifth in 2021 to second in 2025; in the contributed data, at least one CWE in this category was seen in 3.00% of applications. In shops this often means exposed admin interfaces, default passwords, verbose errors, unnecessary HTTP methods, weak TLS and missing security headers.

Using a cloud or packaged e-commerce platform does not exempt you. As the configuration surface grows, a panel that was “finished” can remain open to the internet.

3. Software supply chain failures

A03:2025 Software Supply Chain Failures widens the 2021 “vulnerable and outdated components” category. In e-commerce, themes, plugins, payment integrations and tag managers are often third-party. OWASP notes that this category is less visible in the data but has high exploit and impact scores, and that testing has not fully caught up.

In practice: if you have no inventory of plugins, you also do not know whose code runs on your checkout page. A test should cover dependencies loaded in production, not only the code you wrote.

4. Cryptographic failures

A04:2025 Cryptographic Failures appear in 3.80% of applications in the contributed data. In a shop this can mean weakly hashed passwords, unencrypted storage of card or national-id data, weak TLS, outdated protocols, or a step that still travels in clear text despite a “secure connection” label.

Keeping card data on your own server is a separate burden and, for most shops, an unnecessary risk. A test should confirm where data rest and how they travel. That is not a PCI certification or a QSA assessment.

5. Injection

A05:2025 Injection is a natural part of e-commerce through search, filters, coupons and reviews. OWASP describes this as one of the most heavily tested categories, with a large number of CVEs, ranging from cross-site scripting (high frequency, lower impact) to SQL injection (lower frequency, high impact).

A script in a review can trigger an administrative action in another user’s session. A search parameter can open an entire orders table. Testing should cover both reflected input in the browser and queries on the back end.

6. Insecure design

A06:2025 Insecure Design points to decisions that predate a coding mistake. Examples include a returns flow that leaves stock and payment inconsistent, coupons that stack in unintended ways, a price calculated only in the browser, or a guest checkout that attaches an order to someone else’s account.

These are rarely fixed with a late patch; they come from not doing threat modelling early. Business-logic scenarios in a test should be kept distinct from a plain browser scan.

7. Authentication failures

A07:2025 Authentication Failures show up as weak password reset, session fixation, poor “remember me” cookies and missing brute-force protection. According to Verizon’s 2025 Data Breach Investigations Report announcement, credential abuse was an initial access vector in 22% of the breaches reviewed.

A store login is not “just a login”: order history and saved addresses are personal data. Multi-factor authentication, especially on the admin panel, should be part of the design.

8. Software or data integrity failures

A08:2025 Software or Data Integrity Failures cover unverified update packages, unsigned distribution and untrusted serialised data. In e-commerce, “one-click theme update” can mean unsigned code running in production.

Payment webhooks sit next to this heading: an unsigned notification can create a false “payment received” state.

9. Security logging and alerting failures

A09:2025 Security Logging and Alerting Failures delay the moment you learn about a breach. That matters under KVKK too: Board Decision 2019/10 expects notification to the Board within 72 hours of becoming aware of a breach. It is a mistake to assume that the clock does not run if you never notice; without adequate logs and alerts, you also learn late.

Admin sign-ins, password resets, payment errors and permission changes should be logged, and alerts should actually reach a person.

10. Mishandling of exceptional conditions

A10:2025 Mishandling of Exceptional Conditions is one of the new 2025 categories. Timeouts at the payment provider, stock running out, partial refunds or a double click can leave an order “paid but not shipped” or “cancelled but charged”.

A test plan should include failure and interruption paths, not only the happy path. Otherwise the flaw appears exactly when the application breaks.

How to read a test

These ten headings are not a scan report, and they are not closed by a tool’s raw output. A sound web application penetration test runs with written authorisation, a defined scope and evidence-led findings. Payment, membership and the admin panel should be treated as distinct surfaces. For the wider legal frame, see the guide to penetration testing under KVKK.

Findings are not a guarantee of security or regulatory compliance; they are a starting point. Using a packaged e-commerce platform does not make a test unnecessary; it changes the scope.

Sources