The laser diodes used in eg. CD players have a feedback photodiode, sensing the laser's optical output. If the lasers used for optical fibers have similar mechanism too, and if the diode is sensitive to the light coming to it not only from the chip but also from the fiber itself, and can react quickly enough with high enough sensitivity, maybe it could be exploited. In chosen moments, we could then send a short pulse of laser light into the fiber, then watch the signal from the feedback diode, what gets reflected back from nonhomogenities on the fiber. This would give us the distances of all the splices and connectors, and let us know immediately (if the test is performed eg. once per 5 seconds or with similar short period) that there is an attempt to compromise the line underway. Comparison of snapshots from longer periods apart could also serve to find deterioration of the signal path before it results in failure. The advantage of this approach, if possible, is the ability to add the functionality without having to modify the optical transceivers themselves. It sounds too good to be true, so it probably won't work, but I may be wrong...