“We use Cloudflare” may be the answer when you ask a supplier how its website is protected. It is worth understanding, but it is not yet an answer about your invoices. A service can help deliver pages, filter unwanted requests or challenge automated visitors. Those are different jobs, even when one company supplies them. 8 9 11
The browser is the programme on your device that opens websites. A server is a computer or service that receives requests and sends answers. A request might mean “give me the public help page” or, in a customer portal, “give me the invoice I selected.” For a normal public page, your browser receives files and turns them into what you see. In a protected arrangement, a separate service can sit between that browser and the supplier’s server. 17 11
To see the difference, follow a normal visit. Your browser asks for a page. The request may first reach an intermediary service rather than the computer running the website. That service may return a stored copy of suitable content or pass the request on. Cloudflare is one example of such an arrangement, not the only possible supplier. 11
This intermediary is often called a proxy. It receives a request on behalf of the website and, when necessary, makes a further request to the server behind it. The answer then travels back through the intermediary to your browser. In other words, the browser’s connection reaches the intermediary first; the server behind it is not necessarily the computer directly answering your browser. This extra step explains both the opportunity to filter traffic and the need to protect any other route to the server. 11
Delivering a page and checking a request
A network that delivers content from distributed locations is called a content delivery network, or CDN. Keeping copies of suitable content in that network can reduce the work reaching the original server—the computer or service behind it. This can help the website handle traffic. It is a delivery function, not a decision about which customer is entitled to which invoice. 11
For example, many visitors ask for the same company logo. Once a suitable copy is stored in the delivery network, the network can provide that copy instead of asking the original server to send the same file for every visit. Such temporary storage is called a cache. The benefit comes from avoiding repeated work for content suitable for reuse. I would want a supplier to distinguish this public logo from a private invoice: deciding which content may be shared or stored this way is a separate design decision, not something answered by the letters CDN. 11
A different function examines incoming web requests and applies security rules. It is called a web application firewall, usually shortened to WAF. Cloudflare offers rules chosen by the website operator, sets of rules maintained by Cloudflare, and limits on how often particular kinds of requests can arrive. Which rules are available and active depends on the service plan and settings. 8
Filtering means comparing the request with selected rules before deciding whether it should continue. One rule may look for patterns associated with unwanted traffic; another may limit how many matching requests are accepted in a period. A rate limit is this second kind of control: too much traffic of the selected kind can trigger a chosen response. These decisions can reduce unwanted work reaching the application. They depend on what is configured to match and what action is selected, rather than the brand automatically understanding every business request. 8
A useful comparison is a receptionist checking visitors before sending them upstairs. The check can be valuable, but the accounts department must still decide which records each visitor is entitled to see. In the same way, a request that passes a website filter still needs the application’s own permission checks.
For example, downloading an invoice is an ordinary function. The important question is whose invoice the signed-in person may download. A generic traffic filter is not evidence that the software correctly links that person to the document. I would want the supplier to demonstrate this separately, with invented accounts and invoices in an authorized test environment.
Suppose two customers both sign in successfully and ask for their own invoices. The requests can be ordinary and allowed by the traffic rules. The programme handling invoices must still associate each request with the signed-in customer and return only the permitted document. That decision is called authorization: permission to do the particular thing requested. An additional product can provide additional checks, but the simple observation “a WAF is present” does not demonstrate this customer-to-document decision. 8
Robots are not always unwanted visitors
Some visitors are programs rather than people. Software that visits or exchanges information automatically is often called a bot. Your own accounting software may make perfectly legitimate automated requests. Cloudflare’s bot products can challenge automated traffic, and its documentation explicitly warns that legitimate program-to-program connections or mobile applications may also be affected. 9 10
This is where security and everyday work meet. A person may be able to complete a browser check while trying to sign in. A background connection sending business data may not be able to complete the same challenge. If that connection stops, “the website opens for me” does not establish that the service is working for everyone. 9 10
For instance, the accounting programme sends a request to deliver a test invoice and expects a structured answer saying whether the service accepted it. Instead it receives a browser-check page. The programme was built to exchange business messages, not to behave like a person interacting with that page. The request has not reached its intended result even though a human can still open the homepage. This example explains why the team must check the automated workflow itself and identify which protection changed the response. 9 10
Similar product names can conceal an important difference. Cloudflare’s Bot Fight Mode cannot be skipped using its WAF custom rules or Page Rules. Super Bot Fight Mode does support a narrowly selected exception through a custom rule’s Skip action. That is why “just add an exception” is not a universal instruction: the team first needs to identify the feature that rejected the request. 9 10 13
An exception means that narrowly selected requests do not undergo a particular check. It need not mean that the request bypasses every other check or receives permission to access an invoice. Cloudflare’s documentation describes Super Bot Fight Mode’s Skip action in this way: the request can continue through the remaining security controls. Bot Fight Mode’s different handling of WAF custom rules is why an otherwise plausible solution can have no effect on the problem. 10 13
You do not need to learn the rule syntax. The business question is whether the protection can distinguish the work you intend to allow without simply turning off unrelated safeguards. An exception should solve the particular connection problem, not quietly remove the reason for having protection in the first place.
The route matters as much as the product
An intermediary can only inspect traffic that reaches it. The original server behind it is often called the origin. If that server accepts requests over another route, those requests may not pass through the same filter. Cloudflare therefore describes separate measures for protecting the origin, including restrictions on incoming connections. 11
Knowing a server’s address is not the same as proving that it accepts an unprotected request. Other controls may reject the connection. Nor does a Cloudflare-branded check on one webpage reveal the rules covering every other part of the supplier’s service. The observation describes that visit, not the whole configuration. 8 11
For the same reason, a security review stopped by a browser challenge may not have examined the application behind it at all. The honest outcome is “not assessed through this route,” not “everything passed.” Conversely, failure to recognize a particular protection product is not proof that no protection exists.
A helpful supplier explanation can be short: which pages or services use the intermediary, which checks are enabled, and how intended automated connections continue to work. Keep the promotional website, customer portal and separate invoice-delivery service distinct. Protection visible on the homepage does not answer for all three.
Before a protection change, ask the team to retest real types of work using test data: signing in, viewing an invoice and completing the approved automated exchange. A useful filter should reduce unwanted traffic while leaving those intended journeys usable. That is a more concrete discussion than asking whether a famous brand appears somewhere in the setup.
Sources
[9] Cloudflare: Bot Fight Mode
[10] Cloudflare: Super Bot Fight Mode
[11] Cloudflare: Protect your origin server