You open a business portal, sign in and see your documents. The address starts with HTTPS and there is no certificate warning. That is a useful result: the browser has managed to establish an accepted protected connection. Why would anyone still want to check the settings?
Because that one successful connection does not show every alternative the service might accept, or everything the page might try to load. A modern browser and the service choose parameters for their particular conversation. Seeing that choice is different from checking all the service’s permitted choices. 18
A working connection is one successful route
A browser is the program displaying the portal; the server is the computer system answering it. HTTPS carries their web conversation through Transport Layer Security, shortened to TLS. TLS sets up a protected connection, using secret computer values called keys to make transmitted contents unreadable to an observer and to detect tampering. Before ordinary page data travels, the two programs exchange setup messages and agree compatible settings. 18
A certificate is a signed digital document linking the service name to public information used to check the server’s identity. The browser checks its acceptability for the intended name and the server’s proof that it holds the corresponding secret private key. This identity check is separate from choosing how to protect the conversation. 3 18
For example, your browser offers modern options and the server chooses an option both understand. The visit works. Now imagine an older program that offers only an older option. If the server also permits that option, this second connection can use a different setup. Your first visit did not ask that second question. This is the difference between the settings used for one connection and the full set of settings a service accepts. A reviewer examines the second set rather than inferring it from a single successful visit. 18
A specialist may therefore check which older connection methods the service still accepts. Security guidance rules out obsolete SSL and TLS 1.0 and 1.1. That can matter even when your own browser uses a newer method. The relevant question is what the service permits, not only what your recent visit happened to use. 2
A protocol is a set of rules that computers follow when communicating. Newly designed protocols using TLS must support TLS 1.3 and make it the default; TLS 1.2 may be an additional, non-default option for deployment reasons. That requirement guides new designs. By itself, a report that an existing website accepts TLS 1.2 is not evidence of a break-in. 16
Even within a version, computers can use different methods to establish the secret keys that protect their conversation. Current guidance retires some older methods in TLS 1.2, including what a report may call “RSA key exchange.” That does not mean every certificate with an RSA key is forbidden: checking identity and establishing conversation keys are different jobs. 17
Trying alternative settings involves making connections to the service. It is active testing, not just reading someone else’s public record. Agree the service and permitted checks with the operator before commissioning such an assessment.
Do not repair an identity check by turning it off
A second problem can appear when accounting software connects to a service even though the browser website works. The software may use a different address or have a different trust setup. A certificate must fit the name the connecting program expects. Success at one address does not automatically settle another. 3
Suppose someone suggests disabling certificate verification so the accounting software can connect. That removes part of the check that the connection reaches the intended service. It may suppress the error message without correcting its cause. The appropriate response is to fix the name, trust setup or certificate deployment, not to call the missing check a successful repair. 3
Consider accounting software sending an invoice to api.example.com, while staff use portal.example.com in the browser. The first name may lead to a separately configured service. Its certificate and the program’s accepted issuers must allow the intended identity check to succeed. If the program stops with a certificate error, the operator should identify which condition failed. “The portal opens for me” is not an explanation of that separate connection. Turning the check off would accept a connection without the intended assurance about who is receiving the invoice. Encryption without a properly checked destination does not replace that assurance. 3
This is also why an expiry date alone is not enough. Checking identity involves the intended name and trusted certificate information, not merely finding a certificate whose end date lies in the future. The operator should explain the actual failure in terms of the connection your software uses. 3
A page can request more than its main address
A web page is often assembled from several resources, such as images and scripts — small programs that make parts of the page work. A page delivered through HTTPS can still contain references asking for some resources over unprotected HTTP. This is called mixed content: secure and insecure requests mixed within a page. 12
Suppose the portal page contains instructions to fetch a logo from one address and a script that runs the upload button from another. Receiving the main page does not mean those additional requests have already succeeded. The browser must fetch the referenced parts as well. If the script is requested over HTTP inside an HTTPS page, a modern browser can block it rather than run it through an unprotected transfer. The portal can therefore appear on screen while the upload button fails. Fixing the reference and checking the workflow addresses the real problem; claiming the invoice was intercepted from the reference alone would not. 12
Browsers do not simply treat all of those references alike. They upgrade some requests to HTTPS and block others. Consequently, finding an HTTP reference in page text is not the same as observing that data actually travelled unencrypted. The resource may have been upgraded or may not have loaded at all. 12
For a user, the result can be a page that opens but has a broken part. For a reviewer, the task is to identify what the browser actually did rather than invent a successful attack from the reference alone. A check of the homepage also cannot be presented as a test of every later step in a document-upload workflow. 12
A related setting tells supporting browsers to use HTTPS on future visits. It is called HTTP Strict Transport Security, or HSTS. Once a browser knows and applies this policy, it changes a covered HTTP address to HTTPS before sending the request. If the secure connection has an error, it stops instead of allowing that error to be bypassed.
Compare that with a redirect. There, the browser sends the initial request and then receives a reply telling it to move to HTTPS. The difference is when the change happens: before the request under a known HSTS policy, or after the reply under a redirect. 13
The first visit needs care because the browser may not yet know the policy. A redirect alone does not demonstrate that this initial step is protected in the same way. Nor should an operator extend the policy to all subordinate names without checking them: a rule affecting more services can also break access to services that are not ready for it. HSTS is a setting to understand and maintain, not a magic “secure everything” switch. 13
Faster connections still need sensible application behaviour
When a program reconnects using TLS 1.3, an optional feature lets it send some business data before the full setup exchange is finished. Reports call this 0-RTT or early data. It reduces waiting, but those early messages do not have the same protection against being captured and sent again in another connection as ordinary TLS 1.3 application data. The protocol’s specification explains that limitation. 18
Why can sending earlier matter? A normal setup includes fresh exchanges before business data is processed. The optional early-data shortcut lets a returning program send some data before that full exchange is finished. Under the conditions described by TLS, an attacker can capture that protected early transmission and send it again in another connection. They need not read its contents to repeat it. A server that accepts and processes it again could repeat the requested action. That is the replay risk, rather than a claim that the encryption has simply exposed the invoice text. 18
In plain language, the application must consider whether an early request might be processed again. That matters more for “submit this invoice” than for simply asking to view information. It does not mean every TLS 1.3 connection duplicates invoices. The feature might be disabled or restricted; the operator should know whether it is used for actions that change business records. 18
If that behaviour needs demonstration, it belongs in an authorised test environment with invented transactions. Replaying real invoice requests against production is not a reasonable way for a customer to investigate the question.
Finally, remember which connection was examined. A browser can connect to an intermediary that forwards the request elsewhere. Its successful connection does not independently check the onward link, stored documents or customer permissions. Nor does a result on a supplier’s public website automatically measure its separate invoice service. 18
You need not become the person who configures all these settings. Ask for an explanation tied to the service you use: what was checked, what needed correction and whether the normal workflow still works after the change. “HTTPS works” is a useful beginning. It should not be asked to answer every other question.
Sources
[2] IETF RFC 9325 — Secure TLS Recommendations
[3] IETF RFC 9525 — Service Identity in TLS
[13] IETF RFC 6797 — HTTP Strict Transport Security
[16] IETF RFC 9852 — New Protocols Using TLS Must Require TLS 1.3 (July 2026)
[17] IETF RFC 10015 — Deprecating Obsolete Key Exchange Methods in TLS 1.2 and DTLS 1.2 (July 2026)
[18] IETF RFC 9846 — TLS 1.3 (July 2026; obsoletes RFC 8446)