Do you have any idea exactly what weapon you'd like to incorporate this in? Knowing exactly what one is working towards would help decide things like what kind of force do we have to emulate, and with what rate of fire? Because a Sniper Rifle or a Shotgun are going to have much more kick than the DMR or BR, and will have a longer period of time to afford for resetting their position. It would also let us know what sort of internal space and structure we have to work with. People (not necessarily me) could then start answering your questions with more precision. Keep in mind that because this isn't actually milled out of aluminium and isn't actually detonating explosives inside of it, this isn't going to have the same kick as the actual weapon. This is a good thing, because firing an 8ga shotgun or a 14.5mm high-velocity rifle even once would probably hurt your shoulder.
However, I'm foreseeing your weapon build into two parts, a front and a rear. The rear comprises the butt, stock, and depending on the weapon, may have to extend into the action/reciever. This should, at least in part, be made of some sort of casted rubber or similar material; This will be dense enough to take a beating and not self destruct. You can cast this inside of a pepakura model, but you're likely going to want some sort of soft material (exposed rubber or something else) on the butt. Depending on how the final design works, this may have one or two tubes buried into it. The only difficulty I can foresee is that the UNSC tends to use bullpup firearms (MA5/6 series, BR series, DMR series), and if you want a removable magazine (you do) that's going to take space, hence why knowing exactly what weapon you want is going to be critical for laying out a functional design.
The front half will be the majority of your weapon, with necessary hollow spaces. It will have any electronics you need to use, and probably the majority of the recoil-simulation system. It will have to have a hollow interface with the rear section, and the two should go together and come apart easily - either by friction or, if necessary, thumbscrews. This allows you to easily service or demonstrate the internal mechanism. Because of the way newton's third law works, you may also want to fill this, at least partially, with some sort of absorbtive rubberized compound (also keeping in mind you'll want to preserve some sort of balance of weight).
I think we can safely discard the concept of a gas-operated system, as you have your doubts about whether or not that that would be convention-approved (and if anything goes wrong, compressed gas isn't the safest thing, though the likelihood of injury is actually pretty low). This leaves us with some sort of spring/mechanical system (which would be cheap, but potentially a bit on the complicated side, and I don't see how it can be incorporated into a semi-auto system. It might work very well for a sniper rifle or shotgun) or some sort of electromagnetic/linear actuator system (which might be a bit pricier, but it would tie nicely into any existing electronics systems like ammunition counters, and would work very well for semi- or fully-automatic systems).
Hopeing some of the other people in this thread will pick at my notions of construction and operation.