You know that you would need to calibrate this to each individual, and shock them a few times during the setup. You would need to establish what is normal, what constitutes a lie, and what is the response to being shocked. I hate getting shocked, so I would be fairly stressed out, and anxious. It would be tough to establish a baseline to work from. I'm sure you would want some sort of timer, to prevent repeated shocks when the victim responds to the initial jolt. I doubt that the subject will return to a baseline state, once getting zapped.
Somehow, I can't believe this is for any serious application (maybe interrogation, since Water-boarding has been banned), so figure it must be more of a novelty/gag project, where the truth isn't critical, just shocking the subject for amusement. Most lie detectors measure stress, which will increase with every shock. You'll need to compensate for these changes.