Yes, and I built a very simple one in law school, to illustrate the concept (well to be honest I was an engineer who went to law school as a hobby, and then I took every excuse I could find to build technology things as a hobby within a hobby). Anyway it was just a multimeter, some washers soldered to copper wire and some tape to keep the washers in their place on the subject's hands. When people didn't try to game it, it actually sort of worked, although I have no proof of statistical significance; I just had a feeling of 'better than random'. (our experiments were more focused on how to ask questions to elicit the highest response, as questioning with a polygraph requires some skill at asking questions that are leading or not, and in just the right ways).

Conclusion is still that it's nowhere near reliable enough to be used for anything, but the principle that people show physiological responses when answering truthfully or not about things is sort of true (I don't think that was ever under debate, still was a fun experiment).