Passwords must not contain the user's entire samAccountName (Account Name) value or entire displayName (Full Name) value. Both checks are not case sensitive: The samAccountName is checked in its entirety only to determine whether it is part of the password. If the samAccountName is less than three characters long, this check is skipped. The displayName is parsed for delimiters: commas, periods, dashes or hyphens, underscores, spaces, pound signs, and tabs. If any of these delimiters are found, the displayName is split and all parsed sections (tokens) are confirmed not to be included in the password. Tokens that are less than three characters in length are ignored, and substrings of the tokens are not checked. For example, the name "Erin M. Hagens" is split into three tokens: "Erin," "M," and "Hagens." Because the second token is only one character long, it is ignored. Therefore, this user could not have a password that included either "erin" or "hagens" as a substring anywhere in the password.
( src )
There is a very easy method to detect such 'weak' passwords -- compute Levenshtein distance between password entered and account name, full name, other user's data (if known: D.O.B., address, maiden name, etc).
More about Levenshtein distance.
Yes, I know about these lousy Disqus ads. Please use adblocker. I would consider to subscribe to 'pro' version of Disqus if the signal/noise ratio in comments would be good enough.