You visit a supplier’s website and the Help button leads nowhere. You try a second page, but its instructions contradict the first. Then somebody sends you a report saying the site lacks a protective setting. You naturally wonder: if this is how they look after their website, what will happen to our invoices?

That concern deserves an answer. The difficulty is knowing which answer the observation supports. A public website is part of a business, but it is not a transparent view into every computer, employee and service behind the company name.

What opening a website actually shows you

The program you use to visit websites is called a browser. When you enter a web address, your computer first needs to find where to contact the service associated with that name. It can then ask for a page. A computer system answering that request is called a server. The browser receives information and uses it to put text, pictures and controls on your screen. The page can ask it to fetch more material too. 11

The familiar page is therefore the visible result of an exchange between programs. You have not downloaded the supplier’s whole business. You have received what that page and its supporting services chose to provide for your request. If the page contains prices and a Help button, you can see whether those are present and whether the button takes you somewhere useful.

A supplier can operate a public site that anyone may read and a separate working service where customers sign in to use invoices. Different computers, programs or teams may be involved. Alternatively, the parts may share some components. Their appearance alone does not establish either arrangement. The same logo is a statement of branding, not a diagram of the systems.

This is why the shop-window comparison helps. A broken display light is a real defect and may suggest poor maintenance. It does not let you see whether the stockroom door is locked. A bright, polished display does not let you see that either. The comparison is about the limits of the view, not about excusing the broken light.

A protective instruction belongs to the page it protects

Now return to the report about a missing setting. To understand its importance, we need one more fact about pages. A page can include small programs that run in your browser, for example to make a form react when you select an option. Such programs may be provided with the page or obtained from another source. The browser has to decide what it may load and run.

A website can give the browser rules about which sources it may use for those programs. For example, the contact page needs a program from a source chosen by the supplier. The supplier allows that source in the rules. When the page asks to load the program, the browser checks the request against the rules. If an enforcing rule disallows the requested source, the browser can block the program instead of loading it.

These rules are called a Content Security Policy, often shortened to CSP. They can restrict other security-related behaviour too. They help limit certain ways unwanted content runs, but permission to load a program is not an inspection of its intentions. The protection depends on what the rules actually allow. 5

Imagine that a public page displays a contact form. A report says no such policy was found among the extra instructions accompanying the page. Those accompanying instructions are called response headers. The ordinary visitor usually does not read them; the browser uses them while handling the response. Some policy rules can also be placed within the page itself, with limitations. A check of headers alone therefore supports “not found there”, not automatically “there is no policy anywhere on the page”. 5

There is another distinction. A supplier can use a policy that only reports when rules would be broken, without enforcing the blocks. This can help the supplier see what would happen before changing the behaviour of a live page. It does not provide the same blocking protection. Even if a tool finds the policy’s name, someone still needs to read its settings and mode to understand what it does. 5

You do not need to configure these rules yourself. The important reasoning is that the report concerns how this page instructs the browser. It may identify a worthwhile improvement. It does not, just by naming the improvement, show that somebody can read invoices held in a different service.

The public site can still affect daily work

Do not take the opposite shortcut and conclude that a public site is irrelevant. In our example, the broken Help button already prevents an employee from getting support. No security breach is needed for that defect to cost time. If the site is the normal place to obtain sign-in instructions, it also guides how customers reach their working service.

Suppose an employee reaches the invoice service only through a button on the homepage. If the homepage becomes unavailable, the employee may assume the whole service is down, even though the working service still responds at its own address. A confirmed sign-in address saved as a bookmark gives the employee another known route to try. A bookmark is simply an address saved by the browser so that you can reopen it later without searching.

For this recommendation, “confirmed” matters. Save the address the supplier has established as the correct one, not an unfamiliar result discovered during an outage. Keep the agreed support contact somewhere you can reach without relying on the broken page. These are sensible work arrangements, not a guarantee that a bookmarked service can never fail.

Different systems can also be connected in useful ways. A public page might send a visitor to the sign-in service; the sign-in service might be operated by another team. To judge a security concern, an assessor needs to explain the actual connection relevant to the suspected problem. Being under the same company name is too little information to conclude that a flaw crosses from one part to another.

For example, the rule preventing another customer from opening your invoice belongs to the service that handles that request. After identifying the user, that service must check permission for the particular document and action. A test of that behaviour needs access and evidence beyond reading public prices or contact details. 3

A precise concern is easier to resolve

You can report the broken support route directly: “The Help button on this page does not work. What contact should our staff use, and who will repair the button?” The supplier might explain that an agency maintains the site while another team runs invoice delivery. That is possible, but it does not repair the button. Customers still need a working route to help.

For the protective-setting report, ask for an explanation of what was examined and what the finding affects on that page. A supplier’s answer and a report may differ because they concern different addresses, times or versions of the page. They might also concern the public page versus a page after sign-in. Reporting guidance recommends describing the examined parts and practical limits so readers can interpret these differences. 8

A neglected public site can reasonably make you question maintenance elsewhere. Keep that as a question rather than treating it as a demonstrated relationship between the two systems. Likewise, a well-maintained website is favourable information about the website, not proof that stored documents, customer permissions and recovery all work correctly.

Save the agreed sign-in address and support contact, and ask for the observed fault to be resolved. You can take the shop window seriously without pretending it lets you see the whole business.

Sources

3 — OWASP API1:2023 Broken Object Level Authorization

5 — W3C Content Security Policy Level 3

8 — OWASP WSTG v4.2 Reporting

11 — MDN: How the web works