I just read about the Gettier problem: In my opinion one should better model knowledge as some kind of estimators for probabilities and being justified on some statistical criterion. This should in my opinion avoid the whole Gettier problem (but perhaps introduce some completely different ones?).
I prefer the definition of knowledge given by Orson Scott Card, if I'm not mistaken: something you believe to be true. It doesn't have to be justified and it doesn't even have to be true - you just have to believe it.
I know that the Moon is about 400,000 km away from the Earth. I have never measured the distance and I haven't even tried to check if it's plausible: I read it somewhere and accepted it as such.
Perhaps you could start with this: to formalize any set of laws and criminal procedure similar to the actually existing law, you'll have to define knowledge, and that is a quagmire (http://www.unc.edu/~ujanel/Gettier.htm--you don't have to read the entire paper, but at least read the first section for a feel of how nasty the project gets. If you need more background, here is the description of what the Gettier problem is: https://en.wikipedia.org/wiki/Gettier_problem).