You'd be surprised at how many times a user will type their password in the UserID field. This shows up in a Windows EventCode=4625. By itself it wouldn't be an issue since in a single 4625 event you don't know the userID. However, if you correlate that to other events on the same host, you probably can figure out both the username and password. You'd want to keep the usernames in the field if they didn't look like a password. So my question is this:
Q: How would one propose to selectively find usernames that look like passwords and then substitute a character string (e.g. "********") at index time?
Here are 2 apps that can help:
https://splunkbase.splunk.com/app/3109/ (for fuzzy
matches)
https://splunkbase.splunk.com/app/2734/ (for shannon entropy
)
It depends. What does a username look like? What does a password look like? Can a password look like a username or vice versa?
If usernames are only alphanumeric (other than system accounts that end with "$") and passwords always have a special character in them then you may be able to define a regex that looks for special characters in the username field and masks the field if any are found.
If your usernames allow special characters or you use modern password rules that don't require special characters then what you want to do becomes much more of a challenge.