A Single Meter on a Power Line Can Tell You What's Happening Inside a Building Without Anyone Seeing the Equipment. It Was Invented to Study Dishwashers.

Non-Intrusive Load Monitoring, developed at MIT in the early 1980s to study residential energy use, reads a building's aggregate electrical signature from a single point on the power line to infer what's happening inside -- no access to the equipment required. IAEA nuclear safeguards use a version of the same logic, tracking enrichment facilities' power draw as one input for estimating output. A third, distinct kind of trust from Trusted Foundry and TEMPEST: not who built it, not does it leak -- what does the whole facility's power draw reveal from outside the walls.

A single meter on a single power line, watching nothing but voltage and current, can reveal what's happening deep inside a building without anyone ever seeing the equipment itself. The technique is called Non-Intrusive Load Monitoring, and it didn't come out of intelligence work at all -- it was developed at MIT in the early 1980s by George Hart, Ed Kern, and Fred Schweppe, funded by the Electric Power Research Institute, for a mundane utility-research purpose: understanding residential energy use without installing a meter on every individual appliance in a house.[1] Hart's own early technical reports, in 1984 and then 1985, formalized what he called the Nonintrusive Appliance Load Data Acquisition Method; his 1992 paper systematically established the field.[1]

The mechanism: every device has an electrical signature, whether or not anyone can see it

The method works by watching a home's or building's total electrical draw at a single point -- one digital monitor on the incoming line -- and looking for the specific voltage and current changes each appliance makes when it turns on, turns off, or shifts state. A refrigerator compressor, a water heater, a washing machine motor each leave a distinct, repeatable signature in the aggregate load, identifiable through cluster analysis without ever opening a wall or attaching a sensor to the device itself.[1] The insight scales far beyond a house: any facility running enough distinct, power-hungry equipment leaves a composite signature in its own electrical draw, whether the people running it want that visible or not.

Early 1980sNILM developed at MIT for residential utility research

1Meter needed -- on the incoming power line, not on individual equipment

MTSWU/dayThe unit IAEA safeguards use to convert power draw into enrichment capacity

The same signature, read at the scale of a nuclear safeguards inspection

Nuclear nonproliferation monitoring uses a version of the identical logic, formally, as one tool among several. Gas centrifuge enrichment plants draw a large, distinctive, and remarkably steady electrical load that scales directly with how many centrifuges are actually spinning. International Atomic Energy Agency safeguards inspectors track a declared enrichment facility's average power consumption over the period between material shipments, and use that figure -- alongside physical inspection, material accountancy, and other measurements -- as a basis for estimating the facility's actual separative work output, expressed in metric-ton-separative-work-units per day.[2] It's a genuinely different kind of signal from either of the two disciplines already covered on this site: Trusted Foundry asks who built a specific chip, and TEMPEST asks whether a specific finished device leaks its own signal through the air. Load-signature analysis doesn't care about any single chip or device at all -- it reads an entire facility's aggregate electrical behavior, over the power line itself, from a single point outside the building.

Why does this matter? The same technique that tells a utility a house's dishwasher just started a cycle, developed for reasons that had nothing to do with security, turns out to be a real, standing tool for inferring what a facility is doing at scale -- without needing access to a single piece of the equipment inside it. Trust, in the cluster of ideas this site has been building, isn't one property checked once at one point. It's fabrication trust before a device exists, emissions trust once it's built and running, and now a third kind entirely: aggregate trust, read off the electrical signature of an entire operation from outside its walls. None of the three requires the other two to hold, and none of them is sufficient by itself.

The takeaway A THIRD KIND OF TRUST -- READ FROM OUTSIDE THE WALLS, NOT FROM WHO BUILT IT OR WHETHER IT LEAKS. Non-Intrusive Load Monitoring (NILM): developed at MIT, early 1980s, by George Hart, Ed Kern, and Fred Schweppe -- funded by the Electric Power Research Institute, for residential utility research. One meter, one point on the incoming power line, reveals which appliances are on inside a building -- no access to the equipment itself required. IAEA nuclear safeguards track enrichment facilities' average power consumption between shipments, converting it to metric-ton-separative-work-units per day as one input for estimating actual output. The comparison: Trusted Foundry asks who built the chip. TEMPEST asks whether the finished device leaks. Load-signature analysis asks what an entire facility's aggregate power draw reveals from a single point outside it -- a third, distinct kind of trust.

Sources

georgehart.com / Oliver Parson, Nonintrusive Appliance Load Monitoring — origin and technical method
Nuclear Threat Initiative (NTI), The IAEA's Safeguards System