If you've been in the security community for even a brief time, or you've taking training associated with a certification in this field, you've likely encountered the concept of password hashes. The "Reader's Digest" version of password hashes are that passwords are subject to a one-way cryptographic algorithm for storage, and that same algorithm is applied to passwords that are input, and a comparison is made for authentication. The basic idea is that the password is not stored in its original form.
Now, we 'see' password hashes being collected by threat actors all the time; grab a copy of the AD database, or of Registry hives from an endpoint. Or, why bother with hashes, when you can use NPPSpy or enable WDigest?
Or, if you wanted to maintain unauthenticated persistence, you could enable RDP and Sticky Keys.
Okay, so neither of those last two instances involves password hashes, so what if that's what you were specifically interested in? What if you wanted to get password hashes, or continue to receive password hashes, even across password resets? There are more than a few ways to go about doing this, all of which take advantage of available "functionality"; all you have to do is set up a file or document to attempt to connect to a remote, threat actor-controlled resource.
Collecting hashes is nothing new...check out this InSecure.org article from 1997. Further, hashes can be leaked via an interesting variety of routes and applications; take a look at this Securify article from 2018. Also, consider the approach presented in ACE Responder's tweet regarding modifying Remote Desktop Client .rdp files.
One means of enabling hash leaks across password resets is to modify the iconfilename field in specifically placed LNK/Windows shortcut files, which is similar to what is described in this article, except that you set the IconLocation parameter to point to a threat actor-controlled resource. There's even a free framework for creating shortcuts called "LNKBomb" available online.
Outlook has been a target for NTLM hash leakage attacks; consider this Red Team Notes article from 2018. More recently, Microsoft published this blog article explaining CVE-2023-23397, and how to investigate attempts to exploit the vulnerability. This PwnDefend article shares some thoughts as to persisting hash collection via the Registry, enabling the "long game".
Okay, so what's the big deal? Why is this something that you even need to be concerned about?
Well, there's been a great deal of discussion regarding the cyber crime, and in particular, the ransomware economy for some time now. This is NOT a euphemism; cyber crime is an economy focused on money. In 2016, the Samas ransomware actors were conducting their own operations, cradle to grave; at the time, they targeted Java-based JBoss CMS systems as their initial access points. Over the years, an economy has developed around initial access, to the point where there are specialists, initial access brokers (IABs), who obtain and sell access to systems and infrastructures. Once initial access is achieved, they will determine what access is available, to which organization, and it would behoove them to retain access, if possible. Say they sell access, and the threat actor is "noisy", is caught, and the implant or backdoor placed by the IAB (not the initial access point itself) is "burned". NTLM leakage is a means for ensuring later, repeated access, given that one of the response and remediation recommendations is very often a global password change. If one of the routes into the infrastructure used by the IAB require authentication, then setting up a means for receiving password hashes enables continued access.
What To Do About It
There are a number of ways to address this issue. First, block outbound communications over ports 139 and 445 (because of course you've already blocked inbound communication attempts over those ports!!), and monitor your logs for attempts to do so.
Of course, consider using some means of MFA, particularly for higher privilege access.
If your threat hunting allows for access to endpoints (rather than log sources sent to a SIEM) and file shares, searching for LNK files in specific locations and checking their iconfilename attributes is a good hunt, and something you may want to enable on a regular, repeated basis, much like a security patrol.
For SOC detections, look for means by which this activity...either enabling or using these attempts at hash leakage...might be detected.
From a DFIR perspective, my recommendation would be to develop an evidence intake process that includes automated parsing and rendering of data sources prior to presenting the information to the DFIR analyst. Think of this as a means of generating alerts, but instead of going to the SOC console, these "alerts" are enriched and decorated for the DFIR analyst. This process should include parsing of LNK files within specific locations/paths in the acquired evidence, as parsing all LNK files might not be effective, nor timely.