Surveillance

Your Smart Meter Knows When You Sleep

July 5, 2026 7 min read Haven Team

An old electromechanical meter reported one number a month. A smart meter reports your consumption in intervals of fifteen minutes, sometimes far less, and at that resolution electricity stops being a utility bill and becomes a behavioral log: when you wake, when you leave, when the house is empty, and often which appliance did what. The meter is on the outside of your house, and so is the data.


The technique that extracts meaning from a household load curve has a name and a long pedigree: nonintrusive load monitoring, or NILM, first developed by George Hart at MIT in the 1980s. Every appliance has an electrical signature. A refrigerator compressor cycles in a distinctive rhythm, a kettle is a short sharp spike, an electric oven steps up in stages, a washing machine draws a pattern you can pick out of a graph by eye after ten minutes of practice. NILM automates that reading. Given interval data, it decomposes the single stream of household consumption into individual appliance events.

What that yields depends on the sampling rate. Monthly totals reveal little. Fifteen-minute intervals, the common configuration for billing, reliably reveal occupancy, sleep patterns, and vacations. Research on high-frequency data has pushed much further: at a Chaos Communication Congress talk in 2011, researchers Dario Carluccio and Stephan Brinkhaus showed that with two-second interval data from their German provider, they could infer which film was playing on the household television, because an LCD screen's power draw varies with the brightness of the picture. That result is a laboratory extreme, not a utility's standard practice, but it marks how much signal the wire carries.

The pattern to recognize

This is metadata surveillance in its purest form. Nobody reads your mail or bugs your kitchen; a single innocuous-looking measurement stream, sampled often enough, reconstructs the life behind it. The same logic governs communications records, as we cover in metadata surveillance: the content was never needed.

Who gets the data

The first holder is the utility, which needs interval data for billing and grid management. The questions start with the second hop. Utilities share consumption data with analytics contractors, demand-response programs, and in some markets third-party energy services that customers authorize with a click. Law enforcement has a long history of using energy records: high consumption as an indicator of indoor cannabis cultivation predates smart meters entirely, and those requests were traditionally served with a subpoena to the utility rather than a warrant, because the records belong to a third party.

The legal framework in the United States turned on exactly that point. In Naperville Smart Meter Awareness v. City of Naperville (2018), the Seventh Circuit held that collecting fifteen-minute interval data from homes was a Fourth Amendment search, explicitly acknowledging that the data reveals what happens inside the home. The court then held the search reasonable without a warrant anyway, because the city utility collected it for grid modernization rather than prosecution. Both halves matter: the judiciary has recognized that smart meter data is intimate, and it has still left routine collection warrantless. What protections exist come mostly from state utility regulators; California's privacy rules for smart grid data, which restrict sharing without consent, are among the stronger examples. Whether stored interval data is reachable years later depends on retention rules of the kind we discuss in data retention laws.

Europe ran the argument earlier. In 2009 the Dutch Senate rejected a bill that would have made smart meters compulsory, after privacy objections grounded in the European Convention on Human Rights, and the rollout proceeded on a voluntary basis. The episode is a rare case of a parliament weighing meter granularity as a civil liberties question before deployment rather than after.

The home has other tattletales

The meter measures the whole house, which once required inference to unpick. Increasingly the inference is unnecessary, because individual devices report themselves: smart plugs, connected thermostats, and appliance vendor clouds each log their own slice of activity, a landscape we map in smart home privacy. And a house's activity also shows in its radio environment, which is developing into a sensing channel of its own, covered in our post on Wi-Fi sensing. Energy data stands out because you cannot decline it: the meter is a condition of service, installed and read by a party you do not choose.

What you can do about it

Why this case is worth understanding even if you never touch the opt-out

Smart meter data is a clean teaching example of how modern surveillance actually works. No sensor was aimed at you. A device installed for billing, sampled finely enough and retained long enough, produced a record of your daily life as a side effect. The record sits with a third party, reachable by subpoena, governed by rules written for a one-number-a-month world. Every argument you learn to make about interval data (about granularity, retention, third-party doctrine, and purpose creep) transfers directly to location history, DNS logs, and message metadata. The electricity grid just happens to be the version wired into every home in the country.

Try Haven free for 15 days

Encrypted email and chat in one app. No credit card required.

Get Started →