It's often helpful when security researchers like Guardicore shed light on flaws in Microsoft Exchange - however, the Autodiscover protocol isn't flawed in the way they describe. Even though the issue is hard to replicate, it shouldn't distract from the work you need to do to protect your organization from the underlying reason why people want your credentials.
Lots of excitement was generated when Guardicore revealed a purported vulnerability with the Exchange Autodiscover service. However, the almost total lack of detail about the configuration used for testing and to generate the reported results makes it impossible for Exchange administrators to check the theory against their own deployment. I don't think a problem exists with Exchange Online, but it's possible that poor DNS practice or flawed third-party clients could cause an issue with on-premises servers. The case remains to be proved.