Why the website is an entry point
A company website, an online store, a customer portal or an internal system employees connect to from home are reachable from the internet by anyone, at any hour. Many of them are connected to a database with customer details, orders, documents and sometimes payment details. A vulnerability in such an application is often the shortest path to the organization’s sensitive information, without passing through the office network at all.
A web application penetration test examines the website or system from an attacker’s point of view: how users sign in, what each user can see and do, how the system handles the information entered into it and what happens when something unusual occurs. The test is done with the written authorization of the system owner, within a scope defined in advance and under rules of engagement that set what is allowed and what is not, for example not performing actions that could take down the website or delete data.
OWASP: a shared language for testing and fixing
OWASP is a nonprofit organization and professional community that publishes open methodologies and reference documents for application security. Three of them are especially relevant to a test:
- OWASP WSTG (Web Security Testing Guide): a detailed guide to testing web applications, divided into areas such as authentication, authorization, session management, input handling and configuration. This is the methodology that defines what is tested and how it is documented.
- OWASP Top 10: a list of the most prominent risk categories in web applications, updated from time to time. It is not a complete checklist, but it is a good shared language between the tester, the developers and management.
- OWASP API Security Top 10: a similar list that focuses on APIs, where the risks look somewhat different.
When the report assigns each finding to the matching OWASP category and rates it according to CVSS (Common Vulnerability Scoring System), the developers know immediately what kind of problem it is, and can use accepted remediation guidance instead of developing a solution from scratch.
What actually gets tested
A test according to WSTG covers many areas. These are the main ones:
- Access control: whether a user can see or change another user’s information, or reach administrator functions. Broken access control is at the top of the OWASP Top 10 list.
- Authentication and sign-in: the password policy, protection against repeated password guessing (brute force), multi-factor authentication (MFA) and the password reset process.
- Session management: how the system remembers that a user is signed in, when the session expires, and whether signing out really signs out.
- Input handling (input validation): whether the system checks what is entered into forms, search fields and files, to prevent command injection (injection) and running malicious code in other users’ browsers (XSS).
- File upload: which file types are allowed, where they are stored and who can access them.
- Server settings (security misconfiguration): traffic encryption (TLS), security headers, error messages that reveal internal information and exposed management interfaces.
- Outdated components (vulnerable components): versions of the content management system, plugins and libraries with known vulnerabilities.
APIs and mobile apps
Many modern systems are built so that the website or the phone app is only a display layer, and the real work is done against an API that returns data. What the user does not see on screen may still be accessible through the API, so it is tested directly, separately from the browser.
In the OWASP API Security Top 10, the risk at the top of the list is broken object level authorization: the system checks that the user is signed in, but does not check that the record requested really belongs to that user. The test also looks at returning more data than needed, limits on the number of requests (rate limiting), old versions of the API that stayed active after an update (improper inventory management) and how access keys (API keys) are stored and used.
What automated tools do not find
Automated scanners (web vulnerability scanners) are good at identifying known problems: old versions, missing headers and familiar patterns of input vulnerabilities. They struggle with anything that requires understanding the system. A scanner does not know that one customer is not supposed to see another customer’s invoices, that a discount code is supposed to work only once, or that an order confirmation must first pass a payment check.
Problems like these are called business logic flaws, and they are found only through manual testing by a person who understands what the system is supposed to do. This is one of the reasons a penetration test is not a scan, and why in a grey box test the tester receives users at several permission levels: that way you can check whether moving between customer, employee and administrator is really restricted as it should be.
Which environment to test
One of the important decisions before the test is whether to test the live environment (production) or an identical test environment (staging). Testing the live environment reflects the real situation, but requires care: forms are really submitted, emails go out to customers and orders are recorded in the system. A test environment is safer, provided it is really identical to the live environment in versions and settings.
Either way, the rules of engagement should determine which actions are prohibited, when testing happens and how personal information the tester comes across is handled. Before a test in the live environment, make sure there is a current backup that has been tested by restoring from it, and if the website is hosted with an outside provider, check its terms regarding security testing.
From the report to the fix
Findings from a web test usually divide according to who needs to handle them:
- Code fixes: access control, input validation and business logic. These go to the development team, internal or external.
- Component updates (patch management): the content management system, plugins and libraries. Sometimes an update is enough, and sometimes a component that is no longer maintained has to be replaced.
- Hardening the server and hosting: TLS settings, security headers, file permissions, restricting access to management interfaces, and where needed a web application firewall (WAF).
A WAF can give quick protection against a known vulnerability until the code is fixed, but it does not replace the fix itself. After the fixes, a retest is performed, to make sure every finding was closed and the fix did not break anything else on the website.
Who tests, who fixes and who hosts
A typical website involves several parties: whoever developed it, whoever hosts it, whoever manages the domain and email and whoever is responsible for information security in the organization. When the report reaches each of them separately, it is very easy for a finding to fall through the cracks, especially with server settings that no side sees as its own responsibility.
An IT partner that understands information security and also manages the hosting environment can perform the test according to OWASP WSTG, fix what concerns the server and the infrastructure, coordinate the code fixes with the developers and perform a retest. When the website collects personal information, it also knows how to connect the findings to the requirements of the Privacy Protection Law and Amendment 13, and to set priorities that also take into account regulatory exposure and customer trust, alongside the technical score. A leak of customer details from a website may require a report to the Privacy Protection Authority, and sometimes also a notice to customers.
Summary
- Every website and system that is reachable from the internet and connected to customer information needs a dedicated penetration test.
- Require a test according to OWASP WSTG, with findings mapped to the OWASP Top 10 and the OWASP API Security Top 10 and rated by CVSS.
- Make sure the test includes manual testing of permissions and business logic, beyond automated scanning.
- Decide in advance which environment will be tested, and what to do with personal information in the test environment.
- Divide the findings into code, components and server, and finish with a retest.