A flaw in the design of the Autodiscover protocol used to set-up email clients with Exchange Server can cause domain credentials to be sent in the clear to a server controlled by an attacker.
The Microsoft Autodiscover protocol makes it easier for end users to configure their email client to connect to an Exchange Server – all they need to know is their email address and a password and Outlook with automatically discover all the server details – even if they do not exactly match the domain of the email address. In order for this to work, behind the scenes, the email client necessarily attempts to connect to the Exchange Server using a variety of variations of the domain name found in the email address in search of an Autodiscover.xml file which contains all the settings needed.
According to a report from Guardicore, it is easy for a threat actor to set up a webserver at one of the permutations of server names tried by Autodiscover (or use DNS poisoning to divert traffic to the attacker controller server) and there harvest the credentials included in the connection requests sent from the email client.
A report on Bleeping Computer shows that Microsoft is quickly trying to register in their own right as many of the domain variations that the autodiscover protocol could try, in order to prevent threat actors from exploiting them.
Exchange Admins should, if possible, disable HTTP basic authentication which will prevent credentials being sent to rogue end-points in the clear.