I hold out hope for (f)MRI.
There's some research into guilt measurement with fMRI [1] but it's very hard to test. When my lab tests fMRI thought identification, we need participants to write down exactly what they plan to think about for certain concepts (death, family, etc) and we need them to think about one particular strong memory multiple times. We know, or at least we believe we know, and the participant knows that we know, what they're thinking about. Anything else makes it hard/impossible to classify and won't be accepted in the literature.
Asking someone to pretend to lie in a lab environment, like say responding to "What's your name?" with the wrong answer on purpose, won't elicit the kind of arousal response we want to record like "Did you steal this car?". There's just not a good way to standardize and measure this kind of response without deceiving the participant - you might do some kind of social research on them and ask them questions they aren't prepared for, but understandably no IRB would allow that.