With jpeg(and I think at least some of the mpeg flavors), quantization matrices can be your friend.
Different hardware and software uses different matrices. This isn't a slam-dunk(if somebody just lightened the image a bit to bring out the detail, the quantization matrix would scream "Photoshop!", despite that being pretty innocuous); but it makes it rather harder for a clueless faker to simulate a 'right off the camcorder' "authentic" video if the last compression was almost certainly performed with editing software.
Depending on the details of the format, there are likely to be a variety of other things that are optional or implementation-specific(at least within certain ranges) that can be examined to try to source a given file. If implementation(or quality level/encode settings)-specific details vary between sections of the video, or between parts of individual frames, that's probably a bad sign.
If you have enough footage, and ideally access to the alleged source hardware, you can also attempt to characterize physical defects in the sensor. All digital image sensors, to one degree or another, exhibit imperfect linearity. Some pixels are 'hot', some are abnormally insensitive, this is especially visible on long exposures, or in very dark scenes, where the hot pixels tend to stand out. Onboard image processors have gotten increasingly good at squelching minor sensor noise, so this isn't easy; but a given CCD or CMOS sensor will have a noise pattern that is extremely difficult to replicate. It's just an open question whether you'll actually be able to see enough noise to identify it.