Request a quote
Emergency SectorsCareers About us Blog Get in touch
Close-up of two monitors showing log lines and coloured terminal output

Email filtering for business: what it stops, and what still gets through

Email filtering stops most of the junk, and that is exactly why nobody notices what slips past. The email that costs the most is not spam: it comes from a real address, from a supplier you know, with an invoice that is correct apart from the bank details. This is what a filter can and cannot do, and what you organise around it.

7 min read Written by the ITproposal team
Short answer

Email filtering is good at volume: bulk mail, known malware and senders with a poor reputation. It is poor at targeted email from a genuine, compromised account.

There are three kinds: the filter built into your mail platform, a gateway your MX record points to, and a filter that connects to the mailbox through an API. Each has a different weak spot.

The biggest gap is rarely technical. It is the exceptions list: allowing a whole domain because one invoice got stuck opens the door to anyone who spoofs that domain.

What email filtering stops

An email filter scores every incoming message against a set of signals and adds them up. Above one threshold the message goes to junk, above a higher one to quarantine, and known malware is rejected outright.

  • The sender: the reputation of the sending address and domain, and whether the domain was only just registered.
  • Authentication: whether the message passes SPF, DKIM and DMARC. That tells you whether it really comes from the domain it claims; how to set that up for your own domain is covered in our article on SPF, DKIM and DMARC.
  • Content: patterns that look like bulk mail or fraud, hidden text, suspicious attachments.
  • Attachments and links: known malicious files and, in more complete filters, opening attachments in an isolated environment and re-checking a link at the moment someone clicks it.

That works well for anything sent in large numbers. The filter sees the same campaign at thousands of recipients at once and learns from it quickly.

What still gets through

What a filter sees badly is email that does not look like a campaign. A few forms keep coming back:

  • Email from a genuine account that has been taken over. A supplier’s mailbox is compromised and sends a “new bank account” from an existing conversation. SPF, DKIM and DMARC all pass, because the message really does come from that domain.
  • A domain that is almost right, freshly registered and neatly configured with its own SPF and DKIM records. Authentication passes; it is just not your supplier’s domain.
  • A link to a legitimate service. A shared document in a well-known cloud platform, which only then contains the link to the fake login page.
  • A QR code in an image or PDF, which the filter does not read as a link.

So a filter catches the bulk, not the targeted attack. Changed bank details should always be checked outside email, by phone, on a number you already had. That is a payment procedure, not a filter setting.

Three kinds of filter, three weak spots

The filter in your mail platform. Microsoft 365 includes Exchange Online Protection as standard; Microsoft Defender for Office 365 adds checking links at the time of click and opening attachments in an isolated environment, and is included in some plans. For many small and mid-sized organisations that is a good starting point, provided it has been configured rather than left on default settings.

A gateway in front of your mail platform. Your MX record points to a filtering service, which then passes the mail on. The benefit is a second, independent assessment. The pitfall: your mail platform then sees the gateway as the sender rather than the original source. Without the right connector setting (Microsoft calls it Enhanced Filtering for Connectors) its own checks lose their most important data, and anyone who bypasses the MX record and delivers directly skips the gateway altogether.

A filter connected through an API to the mailbox. It inspects mail after delivery and can pull it back afterwards, including from mailboxes where it already landed. It does not change the mail flow, but it only acts once the message is there.

Which combination fits depends on your licences and on what you already have. How the filter fits with protecting the laptop itself is covered under endpoint and email security.

Quarantine, and the real email that ends up in it

Every filter also holds back email that was fine. The question is not whether that happens, but whether anyone notices before the client calls about the quote nobody answered.

  • Tell users what is waiting for them. A daily digest of held mail works better than a quarantine only the administrator ever sees.
  • Decide who may release what. Users can release ordinary spam themselves; anything flagged as phishing or malware only an administrator, after checking.
  • Agree who may look inside the quarantine. It holds colleagues’ email, with personal data in it. Who may see it and why is a decision for your organisation, not for the provider.

If real email from the same sender keeps going missing, the cause often lies with that sender: a missing or wrong SPF record, or mail sent through a newsletter service that was never authorised. Ask there before you change anything in your own filter.

The exceptions list: where the gap opens

The most common mistake is an understandable one. An invoice from a regular supplier gets stuck, someone adds the whole domain to the allowed senders list, and the problem goes away. From that moment the filter skips every message showing that domain as the sender, including forged ones.

  • Never allow on a domain name or sender address alone. That is exactly what a fraudster fills in.
  • Fix it at the source: have the sender sort out their authentication, or create an exception that also requires the message to pass DMARC.
  • Give every exception an end date and a reason, and review the list every six months. A list without an owner only ever grows.

The same goes for rules users create in their own mailbox. A rule that automatically forwards or moves mail from a particular sender is also a classic sign of a compromised account; have unusual forwarding rules reported.

The report button, and what happens next

Users see what the filter misses. Give them one button to report a suspicious message, Outlook has one built in, and make sure the report lands somewhere that someone actually looks at.

A report is only useful if something happens with it. The person who reported wants to hear whether they were right. And when one user reports a targeted phishing message, you want to know who else received it and remove it from every mailbox. That search should take minutes, not an afternoon.

In the UK you can also forward suspicious email to the NCSC through its Suspicious Email Reporting Service. That helps others, but it does not replace your own handling. What to do in the first hour if someone did click is covered in our article on phishing and the human layer.

Where our work stops

We choose the filter, configure it, manage the exceptions and handle reports the way we agree with you. The filter itself belongs to Microsoft or to the vendor of the filtering product: how good the detection is, is up to them.

We do not guarantee that nothing gets through. No filter does, and anyone who promises it is selling something other than security. That is why a payment procedure that does not rely on email belongs next to it.

After a payment on a fake invoice, recovering the money is work for your bank, and reporting the crime is work for Action Fraud and the police. We help with what the logs show: when the message arrived, who received it and whether an account was taken over.

Frequently asked

Questions we get about this

The five that come up most often.

How does email filtering work?

An email filter scores every incoming message on signals such as the sender’s reputation, the result of SPF, DKIM and DMARC, the content and the attachments. Depending on the score, the message goes to the inbox, the junk folder or quarantine. Known malware is rejected outright.

Is the Microsoft 365 spam filter good enough?

For many small and mid-sized organisations it is a good starting point, provided it has been configured rather than left on default settings. Exchange Online Protection comes with every plan that includes Exchange Online; Defender for Office 365 adds link checking at click time and attachment testing. Whether an extra filter is worthwhile depends on your risk and what you already have.

Why does genuine email end up in quarantine?

Usually because the sender’s authentication is not in order, for example a missing SPF record or mail sent through a service that was never authorised. Sometimes the content resembles bulk mail. Ideally fix it with the sender, and do not add the whole domain to your allowed senders list.

Why does phishing still get past the filter?

Because targeted phishing does not look like a campaign. Email from a compromised account at a real supplier passes every check, and a link to a shared document on a well-known cloud service looks harmless. A filter catches the bulk; against targeted email you need multi-factor authentication, a report button and a payment procedure that does not rely on email.

Can you guarantee that no phishing gets through?

No. No filter stops everything, so we do not promise it. We configure the filter, manage the exceptions and handle reports. If a fake invoice has already been paid, recovering the money is work for your bank; we provide what the logs show.

Related services

Further reading

What email filtering connects to.

Shall we look at what your filter lets through?

Show us the settings and the exceptions list of your mail filter. We will tell you which exceptions are a risk and what is missing from the configuration.

Practical IT knowledge in your inbox

New guides on management, security and the workplace, written by the people doing the work. No sales talk, and you can unsubscribe in one click.

We use your address for the newsletter only. See our privacy policy.

Request a quote Call