You can open your mailbox through an encrypted website and still have a separate question: how was the email protected before it reached that mailbox? The connection from your browser to your mail provider is not the connection from the supplier’s mail service to yours.
Your browser is the program opening the mailbox website. A mail server is a computer running software that receives, stores or sends email. The supplier’s service can hand a message to your provider’s server before you open the browser. When you later read it, the browser retrieves information from your provider. There are therefore at least two different communications to consider: the delivery to the provider and your access to the mailbox afterwards. 10
Email can pass through several systems. Protecting one connection does not automatically protect the next, and the services handling the message can still process it. This is why ordinary server-to-server email encryption should not be confused with end-to-end encryption that keeps the contents from the intermediate services. 10
Protecting one handover
Picture a delivery route with several legs. Arranging protected transport on one leg tells you something useful about that leg, not about every warehouse or vehicle involved. Email connection protection works with the same kind of boundary.
Encrypting a connection means transforming its traffic so someone merely watching the path cannot read the message in its ordinary form. The two services communicate using the protected connection and can recover the information at their ends. This helps with exposure during that transfer. It does not prevent the receiving service from storing a readable message or making it available to an authorized mailbox user. The protection ends at the connection’s ends. 10
Mail servers can use TLS, the technology that encrypts a connection and supports checking the other end’s identity. Mail servers exchange messages using rules called SMTP. A command called STARTTLS lets them switch that connection to TLS. But supporting encryption is not the same as requiring an appropriately verified encrypted connection every time. The systems need rules for what to do when the protection is unavailable or cannot be checked. 10
Without a requirement, a sender may have to decide whether to deliver despite a TLS problem. An attacker interfering with the connection can also try to make encryption appear unavailable. STARTTLS by itself does not supply a universal policy that every sender must use for every destination. 10
For a business receiving invoices, the useful question becomes: can our domain tell other mail services which receiving servers are acceptable and what protection they must use?
Publishing the receiving instructions
A mechanism called MTA-STS does that for senders that support it. The domain publishes instructions identifying its acceptable mail servers and a policy for TLS protection. In enforcement mode, a participating sender applying the policy must not deliver through a receiving server that fails those requirements. The message may have to wait for a retry, and persistent problems can lead to failed delivery. 8
For example, your business receives invoice notices at [email protected]. The domain, the name after @, is example.com. Its receiving mail service has its own server name. A supporting sender learns the receiving rules for example.com, checks that the server selected for delivery is among the permitted names and checks the protected connection. The receiving server presents a certificate: information used to check its name and its connection identity. Under an enforced policy, merely offering some encryption is insufficient if the required certificate checks do not succeed. 8
Suppose the receiving provider replaces a certificate incorrectly and the name no longer matches. The invoice itself has not changed. The sender has a transport problem: it cannot complete the required check of the destination. Stopping that transfer is the intended effect of enforcement until a suitable delivery can be made. Otherwise, an attacker presenting the wrong destination could benefit from the sender treating a failed identity check as unimportant. 8
The instructions have two public parts. First, a small entry in DNS, the Internet directory for domain information, announces that a policy is available and gives its identifier. The sender uses this announcement to learn that there are receiving instructions to fetch.
Second, the sender retrieves the full policy as a text file through HTTPS, the protected form of a web connection. This lets it check the policy service’s identity and protects the file during retrieval. For our example domain, the dedicated address is https://mta-sts.example.com/.well-known/mta-sts.txt. It is a technical policy location, not the company’s homepage. 8
The policy gives its mode and how long a sender may keep a copy. In testing and enforcement modes it also lists permitted receiving-server names or name patterns. An administrator must make these instructions agree with the receiving service actually in use. Finding the DNS entry alone does not establish that the policy file is accessible or correct. 8
Direction matters here. Publishing MTA-STS for your domain describes how supporting services should deliver mail to you. It does not by itself establish how your provider handles mail you send to everybody else. For that, your outgoing service needs to support and apply the relevant recipients’ policies. 8
Why testing and remembered policies matter
MTA-STS has a testing mode as well as an enforcement mode. In testing mode, these policy failures do not require the sender to stop delivery; they can instead be reported. A domain can therefore learn about a problem before making it a reason to withhold delivery. A third mode, none, indicates that the domain is not applying MTA-STS protection. Testing should not be reported as though it already enforces the protection. 8
Now consider moving your business mail to another provider. Some senders may still hold a copy of your previous policy. If that policy permits only the old receiving servers, the new service can fail its requirements. The mail-routing change and the policy change therefore belong in the same migration plan. Deleting a record does not instantly erase policies already remembered elsewhere. 8
Remembered policies also explain part of the protection. Once a sender has learned a policy, it can use the saved copy. At the very first discovery, however, interference can prevent the sender from learning that a policy exists. MTA-STS improves protection, but does not make every first contact immune to interference. 8
The reverse conclusion would also be wrong: no MTA-STS does not mean all email travels without encryption. Mail services can use TLS without it, and other transport-security arrangements exist. The missing mechanism tells us about that mechanism, not the protection of every actual delivery. 8
How the provider notices a problem
A related mechanism, SMTP TLS Reporting, usually shortened to TLS-RPT, lets participating senders report transport-policy successes and failures. The receiving domain publishes a separate DNS entry specifying where those reports should go. The reports can help reveal certificate problems or a mismatch between the policy and the receiving server. 9
A report is a diagnostic clue, not proof that somebody intercepted an invoice. The cause could be a configuration mistake or a possible attack. Likewise, an empty reporting mailbox is not proof that every route worked: reports depend on participating senders and successful reporting. Someone needs to interpret what arrived and notice whether useful feedback is arriving at all. 9
In our certificate example, a useful transport report could tell the receiving provider that senders are encountering a certificate check failure. The provider can compare the report with the current server settings, repair the mismatch and look for later successful observations. The report has helped connect a delivery symptom with a setting to investigate. It has not disclosed the invoice’s commercial contents or identified an attacker. 9
If your business changes mail providers, ask who will coordinate routing, receiving-server names, certificate readiness and the published policy. These pieces describe the same receiving service from different directions. A working mailbox login tests your access; it does not test all of those delivery arrangements. A planned check using harmless example messages can exercise delivery without circulating a real customer invoice.
For the business owner, a clear explanation might be: “We publish an enforced receiving policy, our provider checks transport reports, and a mail-provider change includes updating that policy.” It describes an operating arrangement rather than just another acronym on a sales page.
None of this checks whether an invoice is correct. These mechanisms concern the connection between mail systems, not who can read a stored mailbox or approve a new payment destination. When an invoice notice is delayed by a transport problem, use your established route to the underlying document rather than weakening protection merely to make the notification arrive.