Any modern light water reactor (LWR) can be used to create fuel for nuclear weapons, and many are. In the interest of preventing nuclear proliferation, it would useful to have a sensor independent of the reactor that can measure if the reactor is being used to create fuel for weapons. The fission process gives off a large number of anti-neutrinos.  It is impossible to shield or hide this neutrino flux and their detection can provide a measure of the composition of the nuclear fuel in the reactor. Armed with this information, third party inspectors could monitor the use of the reactor. This paper will describe the basic mechanics of this measurement and the associated problems that make this endeavor ineffective.
Each fission event releases 6 anti-neutrinos as the initial neutron-rich heavy element decays into lighter elements and fewer total neutrons. Neutrinos have an extremely small interaction cross section with ordinary matter and thus freely escape the reactor. These neutrinos can be detected external to the reactor via a reverse beta-decay process in which an anti-neutrino and proton are converted into a neutron and positron. However, because the products have greater rest-energy than the reactants, the neutrino must carry a minimum initial energy for the reaction to proceed.
Inside a LWR, Pu-239, the most common fuel for weapons, is continuously created as U-238 absorbs neutrons. One of three things can then happen to this Pu-239, it can fission, it can absorb another neutron to become Pu-240, or it can be removed from the reactor. As the reactor burns through its limited initial supply of fissile U-235, Pu-239 will build up and contribute significantly to the energy output of the reactor as it too undergoes fission.
The trick for monitoring reactors using neutrinos comes from the slight difference in fission process of U-235 versus U-239. On average, the neutrinos released from a U-235 fission are more energetic than those released from a Pu-239 fission. Thus a neutrino from U-235 is more likely to be above the threshold energy for reverse beta-decay and thus more likely to be detected. The rate of neutrino detection is given by
where γ is a proportionality factor that accounts detection efficiency, solid angle to the reactor, etc., and k(t) is factor that accounts for the changing proportion of U-235 and U-239 in the reactor.  Assuming that the reactor power, Pthermal, remains constant, changes in the reactor composition could be inferred from changes in the neutrino detection rate. If the Pu-239 was removed from the reactor rods, and the reactor restarted at the same thermal power, this could in principle be detected and reported.
This technique has been successfully experimentally demonstrated at the SONGS reactor.  The NUCIFER detector is a design that has been proposed with the intention of being portable and easily operated by technicians who are not neutrino experts. The size of the proposed device is relatively small but still 15 m3. 
While possible in principle, such a device has flaws that make it ineffective as a tool for anti-proliferation. The fundamental problem is that it is far easier for a rogue state to stay one step ahead of the inspectors than the other way around. Say the IAEA had NUCIFER detectors ready for deployment. In what easy way could the rogue state make NUCIFER detectors useless? The Pthermal of the reactor could be varied with time. Or, there could just happen to be no convenient place anywhere close to the reactor to put a 15 m3. This is by no means an extensive list of foils. A creative reader could dream up many more.
Reactor neutrino detectors are a far less attractive tool for monitoring reactor composition than advertised. The complicated nature of the process provide many opportunities for the measurement to be contaminated or rendered impossible to make. In addition, the statistical and abstract nature of the measurement make it inaccessible to everyone without a physics degree. Any public discussion of the results could easily be devolved into jargon indecipherable by any public servant. However, there is a broader problem. A fancy measuring device is not necessary to expose nuclear weapons ambitions. One only needs to don a common sense cap and ask, would Iran build reactors for peaceful energy purposes while simultaneously burning away their stranded natural gas? No. The larger problem then is not outing reactors used for breeding fuel, but determining how to stop them. It is possible that good reliable third party measurements on the reactors in questions would aid this process. Neutrino detectors, with their fog of complications, cannot serve this role.
© David Berryrieser. The author grants permission to copy, distribute and display this work in unaltered form, with attribution to the author, for noncommercial purposes only. All other rights, including commercial rights, are reserved to the author.
 V.A. Korovkin, "Measuring Nuclear Plant Power Output by Neutrino Detection," Atomic Energy 76, 712, (1994).
 A. Porta et al., "Reactor Neutrino Detection for Non-Proliferation With the NUCIFER Experiment," IEEE Transactions on Nuclear Science 57, 5 (2010).
 N. Bowden et al., "Experimental Results Rrom an Antineutrino Detector for Cooperative Monitoring of Nuclear Reactors," Nucl. Inst. Meth. 572, 2, (2007).