Detect Multiple Account Numbers Using Regex Groups - regex

Question

My SVP wants me to update our regular expression rules on our email system to better detect US bank account numbers. The issue is that bank account numbers in the US are not standardized, they can be between 6 and 17 digits.
We currently use qualifying terms to detect specific strings that we have identified as needing to be blocked. Our current rules are variations of this:
(?i)bank\saccount\s[0-9]{6,17}
The issue that I need to solve is the need to detect the numbers even if they are prepended or followed with bank account. I know I can find a single example with this:
(?=.*?(bank\saccount))(?=.*?(\d{6,17}))
But my SVP also wants to be able to detect the number of account numbers in a particular message. I've tried adding a third capture group with a greedy quantifier so that it grabs a different number than the second:
(?=.*?(bank\saccount))(?=.*?(\d{6,17}))(?=.*(\d{6,17}))
Here is a sandbox with a couple of examples:
https://regex101.com/r/hqIEaR/3
I am new to regex, is there a way to set up this expression to return a number of matches equal to the instances of 6-17 digit numbers in a message where the string "bank account" is present?

Answer 1

Maybe simpler is better:
(?<=\D|^)\d{6,17}(?=\D|$)
Test here.
The idea is that you find all numbers with 6..17 digits. They are probably account numbers.
The problem is that looking for "bank account" is useless. Your statement is:
The issue that I need to solve is the need to detect the numbers even if they are not prepended by "bank account ".
So if that string may or may not be there, just ignore it completely.
How can you differentiate between an account number and a SSN? That is the topic for another question.
If "bank account" AND the numbers must be found, but with no clear relationship between them (considering their location in the text), I would actually use two searches:
a search for bank account;
If the first search succeeds, a second search for the numbers.
I expect (no proof) that it will be even faster than doing it entirely in regex, since many things will not be done at all.

Answer 2

Since you are using PCRE compatible engine, you may use a regex like
(?is)(?:\G(?!\A)|\A(?=.*\bbank\saccount\b)).*?\K\b\d{6,17}\b
See the regex demo.
(?is) - case insensitive and singleline/dotall modes on
(?:\G(?!\A)|\A(?=.*\bbank\saccount\b)) - either the end of the previous match or start of string (\A) that has a blank account whole word anywhere to the right of the current location (a (?=.*\bbank\saccount\b) positive lookahead)
.*? - any 0+ chars, as few as possible
\K - match reset operator that discards the text matched so far from the overall match memory buffer
\b\d{6,17}\b - 6 to 17 digits matched as whole words (no other letters, digits or _ chars can appear on both ends).