Sincerely i see many problem on the implementations of your idea especially whereas we speak of a critical area of the system where even in cases you being able to make something functional, because you should even consider system optimization that in this area are essential
(the software operates - practically - in real time).
I think a good start could be:
understand the source code of netfilter;
read many book about the various system available for doing what you want (generally a lot complicated) and often there are only concept, idea, without real implementation available;
very good knowledge of c language;
ask for help in places frequented by kernel developers (for example mailing list after having made some piece of code) because is improbable you could obtain help in a forum about questions so difficult;
a lot of time
After all that, if you are sufficient able to realize something, you should see if the performance (essential in this application) are sufficiently good or not.
I understand your enthusiasm and I appreciate indipendently how "practical" it is your idea.