There seems to be an increase in temporary SPF and DMARC failures, affecting both legitimate mail and messages spoofing internal addresses. Some spoofed messages are reaching inboxes. What should I check to figure out what's going on?
Seeing more SPF and DMARC temporary failures in Exchange Online?
Answers
Start by separating temporary errors from actual authentication failures. An SPF temperror can mean a DNS lookup timed out or returned a temporary failure. It doesn't by itself mean the SPF record is invalid.
Compare headers from several affected messages and note the sender domains, times, sending IPs and authentication results. If unrelated domains are affected around the same time, that points toward a possible DNS resolution issue. Successful lookups from your own test hosts don't rule out intermittent failures along another resolver path.
Reply to this answer
For the spoofs that landed in the inbox, check the Authentication-Results header, including compauth, and trace the route into Exchange Online. If a third-party gateway sits in front of your MX, check whether Enhanced Filtering for Connectors is configured correctly. Otherwise, filtering may evaluate the gateway IP instead of the original sender.
Also look for allowed sender entries and mail flow rules that bypass spam filtering. An inbox delivery could involve more than the SPF result alone.
Reply to this answer
Check how the tenant handles temporary authentication errors before treating a delivery as proof of a new exploit. A temporary error is not the same thing as a definitive spoof verdict, and acceptance or deferral behavior can affect what happens next. It may also be worth investigating whether the spoofed internal-address messages are using direct send.
Reply to this answer
Be careful about switching straight to broad quarantine rules. If legitimate internal mail is already showing spoof-related errors, test the effect on a sample first and work out which flows need fixing or carefully scoped exceptions. Otherwise you could swap an authentication problem for a lot of disrupted mail.
Reply to this answer
Pull the relevant log fields into a CSV if the raw traces are getting unwieldy. Grouping by time, sender domain, source IP, authentication result and delivery action can make it easier to spot whether this is one mail flow or several unrelated ones.
Reply to this answer
Add your answer
Answers are reviewed before publication.
_d83!mivlori.2:
Keep the headers and message trace for those deliveries. They should help distinguish a direct send path from mail that came through a gateway.