For a display, have an agreed-upon, and constantly changing, pseudo-random pixel/row/column update order. This would propagate up to the actual row/column selectors. This would mean that the signal couldn’t be observed except at the actual output of the display row/column selector, and would have to monitor each individual to grab the video data. This would mean monitoring and post process a huge number of signals to get a raw image.
Maybe this is how it’s done.