Data Mining Home Water Usage; Your Water Meter Knows You A Bit Too Well

The average person has become depressingly comfortable with the surveillance dystopia we live in. For better or for worse, they’ve come to accept the fact that data about their lives is constantly being collected and analyzed. We’re at the point where a sizable chunk of people believe their smartphone is listening in on their personal conversations and tailoring advertisements to overheard keywords, yet it’s unlikely they’re troubled enough by the idea that they’d actually turn off the phone.

But even the most privacy-conscious among us probably wouldn’t consider our water usage to be any great secret. After all, what could anyone possibly learn from studying how much water you use? Well, as [Jason Bowling] has proven with his fascinating water-meter data research, it turns out you can learn a whole hell of a lot by watching water use patterns. By polling a whole-house water flow meter every second and running the resulting data through various machine learning algorithms, [Jason] found there is a lot of personal information hidden in this seemingly innocuous data stream.

The key is that every water-consuming device in your home has a discernible “fingerprint” that, with enough time, can be identified and tracked. Appliances that always use the same amount of water, like an ice maker or dishwasher, are obvious spikes among the noise. But [Jason] was able to pick up even more subtle differences, such as which individual toilet in the home had been flushed and when.

Further, if you watch the data long enough, you can even start to identify information about individuals within the home. Want to know how many kids are in the family? Monitoring for frequent baths that don’t fill the tub all the way would be a good start. Want to know how restful somebody’s sleep was? A count of how many times the toilet was flushed overnight could give you an idea.

In terms of the privacy implications of what [Jason] has discovered, we’re mildly horrified. Especially since we’ve already seen how utility meters can be sniffed with nothing more exotic than an RTL-SDR. But on the other hand, his write-up is a fantastic look at how you can put machine learning to work in even the most unlikely of applications. The information he’s collected on using Python to classify time series data and create visualizations will undoubtedly be of interest to anyone who’s got a big data problem they’re looking to solve.

68 thoughts on “Data Mining Home Water Usage; Your Water Meter Knows You A Bit Too Well

  1. Polling by second can tell a lot of information. When our town polls for billing purposes I think it’s more like once an hour or even once a day which now blends all of that usage together and makes it harder to find patterns.

    1. But the point is that any average Joe (average Joe with RTL-SDR experience anyway) can learn all this information about what goes in inside your home. Not that the utility company cares how many times you flush the toilet.

      1. The current “state-of-the-art” smart meters encrypt their radio signals, so no, the average Joe cannot hijack the data. But there are some meters that don’t encrypt and even more scenarios where the utility may deploy them without utilizing the encryption feature….with some brands the key management can be a real hassle.

        1. I would be shocked if any significant percentage of radio-enabled meters in the wild are actually encrypted. Shocked. Any data about their adoption, or as you mention–how often the security is turned on?

      2. Are you kidding, in the future the government will monitor your water usage and send you off to the gulag if you shower too long (more than 30 seconds) or flush your toilet more than once a day, for the environment, of course.

      3. Any average joe can also watch your house with his eyes and find out how many kids you have. If someone’s got to sit with an RTL-SDR, presumably within a few 10s of metres of my house, they can learn a lot more by watching with a camera or binoculars.

      4. I am professionally fairly familiar with the water meter market.

        Second by second monitoring is not a commercially implemented service anywhere in the world that I am aware of. Both economics and physics get in the way.

        SMART Water meters are almost all battery powered and designed for a life of 6-10 years. Battery technology simply does not support that measurement resolution (before we discuss the cost of communications)

        At this stage in thier development, meters are typically contracted to be read daily with 30 or 60 minute consumption data returned.

        All reputable mftrs have some form of encryption in place, although given the requirement for battery power, this is often fairly simplistic.

    2. Polling by the second requires specialized hardware. Many smart meters can’t even do it and those that can need to be put into diagnostic mode which requires special hardware and software. The referenced study likely used a dedicated data logging device. Smart meters typically record readings hourly or monthly, depending on how they’re being read (drive-by or AMI).

  2. “yet it’s unlikely they’re troubled enough by the idea that they’d actually turn off the phone.”

    Doesn’t help much if the alphabet agencies can listen in with the phone off :P
    Probably why all late generation phones have non-removable batteries :D

    1. ” We’re at the point where a sizable chunk of people believe their smartphone is listening in on their personal conversations and tailoring advertisements to overheard keywords, yet it’s unlikely they’re troubled enough by the idea that they’d actually turn off the phone.”

      All these security researchers can find everything else, but they can’t find a spy in the house.

  3. So, houses in his area don´t have a per-house water tank ? That means, when *whatever* happens with the street plumbing, they get no water until the problem is repaired ?

    1. Wait, where do you live where houses have their own water tanks? I’ve never heard of this in the US. Beyond the hot water, anyway. Though even some homes are going away with that now.

        1. Jace, are you sure?
          Many tanks logic units start to refill only after a given low level threshold, witch a constant flow. So they mask all the small utilities specific patterns.

          1. The classic US water heater uses the pressure of the incoming water to push out the hot water. There is no fill valve.
            It sounds like water infrastructure in other parts of the world is very different.

      1. Sure. Normally about 0.5 or 1 m^3. Very big houses or places with too many people ( schools, hotels, etc ) use bigger ones, or multiple.

        So, if something happens in the street plumbing ( like some digging gone wrong ) , or some problem with the pumping ( once during a flood the place where water is pumped out of the river was under water, and they took two or three days to normalize operations ) , then people can have a minimum of water for the most necessary tasks .

        And that irks me, when people living in the highest parts of cities appear on TV complaining about only having water two or three days on a week, and having to store it in buckets when it comes. Small tanks are cheap, and are even suggested in the manuals of the water company.

          1. Nowadays, Brazil. And have seen the same kind of installation in countries around here, also some in Europe ( Portugal, Spain, I believe some in Germany also. But since I was just on vacation , wasn´t going to dedicate much time to study construction techniques of other places .. ;)

      2. It’s extremely common in the U.K. most houses have hot and cold water tanks, unless you’ve switched to a condensing combi-boiler.
        But water failures – even low water pressure – in the U.K. is extremely rare, and they’re very quick to deploy emergency supplies.

        No tank has the benefit that all taps are fresh drinking water (if you’ve got a tank, usually only the kitchen tap is fresh mains, and tanks can get pretty nasty).

      1. But then, whenever something bad happens ( loss of power, broken piples ) people have to endure with no water.

        Was going to use Puerto Rico as an example, but that is an island and one very different can of worms.

        So, just considering those places that get affected by hurricanes or forest fires in the US : when the transmission lines get hit, people served by them , even if not directly affected, have much problems because of the loss of those “buffers” in the water system …Loss of power / internet can be managed ( well, unless you are a cellphone-addicted teenager ) , but water is in a more basic level

        1. You are correct. I have seen neighborhoods where each house has its own water storage tank in my travels to Peru and Venezuela, but they are practically unheard of in those parts of the USA served by municipal water supplies (which is almost all but the most rural parts of the country). We do, however, have large elevated storage tanks as part of our municipal water systems. In the plains states, each small town has a water tower. In more mountainous areas, the water tanks are on hillsides above the neighborhoods. Either way, there is enough stored and pressurized water to last a few days. If electrical service is out for longer than that, generators can be brought in to power water pumps to refill the tanks.

          I have lived in various parts of the US for over 50 years, and never experienced a disruption of municipal water service.

          1. Wow, you’ve never experienced a disruption to the water service? That is impressive.
            I’m in my 30s and have lived most of my life in the UK, I’ve experienced water that we were advised to boil before consuming or washing with, and a complete absence of water when something broke the water main serving the side of the town I lived in. In the latter case service was resumed, but at a lower than normal pressure, after only a few hours and normal service was back by the next day.

      2. As one who has used a well for almost two decades and know many others who do as well I can tell you that we do not have water storage tanks. We have a pressure tank but that doesn’t count as like most water heaters it is filled most any time water is used other than a hand wash as most pressure switches attached to the well pump kick on after a 20 psi pressure change. Sounds like a lot but flushing the toilet will set it off.

    2. In 40 years I have only seen a handful of water outages. I could probably count them on one hand if I could remember them. All lasted less than a day with maybe a boil advisory for the following day. Until a couple of years ago. We had a do not drink advisory for a whole 2 days due to toxic algae.

      I was lucky and missed most of that 2 days because I was away camping. Also fortunate, someone called me and let me know it was happening. Before I left I filled my water bag from the campsite water supply, a well with a hand pump. That’s kind of a hard thing to fail. I actually avoided the water an extra day because I had plenty and I figured better safe than sorry.

      If it happens again I’m not worried. I commute to a different town to work. It’s water comes form a different source. I can bring my camping water bag with me and fill it there! If somehow both failed simultaneously (has never happened to me) I could go to relatives houses with their own wells and generators. Or… just go camping!

      And what if I did have my own tank? Wouldn’t I have to take care to make sure bacteria and other nasties don’t grow in it? Worse, when the next do not drink advisory hits will it not contaminate my tank before I get the notification and turn off the supply? It sounds like a whole lot of wasted effort, money and space to me.

  4. And here I sit in a house that has the same old water meter since probably sometime in the 70’s…(if not older)
    Though, one can check current water usage if one stares at it for a while. (the lest significant digit on it is still in hundreds of liters….)

    But though, I kinda live in a place where water isn’t even remotely expensive.

    Like the old mechanical meter I have gets checked maybe once a decade if not less often then that…. (So the resolution of that consumption data is fairly poor.) There is likely another meter somewhere in the larger water system, metering the consumption of the local area in a more real time fashion….

    Though, the water company does send out leaflets once in a while asking what the water meter currently states, and the numbers should add up if people are honest. (though, this doesn’t even happen once a year…)

    But smart water meters is kinda something that alludes my curiosity, how expensive does water need to get for it to be worth while having such a meter on each resident?

    1. Water in my part of North Carolina can be $50-$100 a month for a small, careful household. The cities like to read the meters monthly. Smart meters let them do that without physically looking at each meter once a month.

    2. My water (midwest) costs about a 1/3rd of a cent per gallon, though if you include the sewer costs it’s about an 1/8th of a cent per gallon.

      Metered individually, and a meter reader walks by once a quarter with a RF reader…

      Sewer is billed based on the water use during the quarter that happens in the winter… so my dec/jan/feb water use determines my sewer bill for each quarter of the the next year. (presumably all the water in the winter goes down the drain, and none of it is used to water the grass (snow))

      I usually end up using 8-10k gallons of water a quarter…

      When I purchased my house in the spring (foreclosure) the pipes had frozen and burst over the winter (pipes were nice enough to burst over the floor drain no water damage) so I got billed for an incredible amount of sewer use that year compared to my water use…

      I hear they are working on replacing all the meters in my area, but no one has reached out to me about it (I suspect they want more wireless range so the meter reader can just drive past from the road) given that the meter is powered by the flow of water I don’t suspect it will have enough power to broadcast high resolution water usage…

      1. I don’t know which meter they are using, but the Sensus meters are powered by a battery sealed inside the meter. I’m sure they have to limit transmitting to hit the 20 year life goal for the battery.

      2. I still wonder how it gets affordable…
        Since if I understand you right, you pay 1/3 of a cent per gallon, and consume roughly 9K gallons a year.
        I get that to be roughly 30 dollars a year, hardly a lot of money.

        Like the water company could just charge a flat rate for maintenance of the pipes and such…. Though, there might be a service fee on top of the cost of the water itself, but the water is practically free. (For the average household that is, farmers, industry, stores and offices is a different story.)

        So I think it seems a bit silly to put in a meter to measure something that is almost free…
        But if it lasts 20 years, then that is though $600, but that is still not much.

        If water costed a few times more, then I can understand the need to measure it, but for straining out a few dollars a year out off a household seems silly to put so much effort into… (Though, it could just be a fee to stop people from wasting excessive amounts of water just because it is cheap.)

    3. “But smart water meters is kinda something that alludes my curiosity, how expensive does water need to get for it to be worth while having such a meter on each resident?”

      When the world in general starts having more clean water crises.

  5. I discovered this and more when I built my LAN of things for the off-grid homestead. Holy Moly, there’s a lot of info about what you’re up to from things you might not suspect at first.

    I had, as part of the water system, which is based on rain collection, inflow, outflow, temperature of a couple spots likely to have freezing issues in the winter, temperature of the basement/crawl space, a barometer here and there (because the 180 or 280 chips give you that)…the usual stuff, more or less.
    Since I live here and know what I’m doing or just did, looking at the plots (I use a raspberry pi or few to serve MySql, NGINX, gnuplot etc) – holy crap – you can not only see the toilet flush (someone’s home!) via the basement temperature alone (the pipes get cool suddenly), but also – how well they washed hands after. Kitchen work has its own signature patterns.

    Barometers are great LF microphones. Open a door or window…or just breeze by…you get the idea. Even humidity…cooking, showering.

    Then there’s the solar power system and weather stations…power use patterns are pretty informative.

    I’m sure you could get more out of correlations if you tried – in my case I didn’t need or want that level of resolution and while I did (and do) sample fairly fast, it’s merely to use averaging and other DSP tricks to get better data, which is then reported at longer intervals, precisely to hide this kind of thing…well, it makes prettier plots too…and saves room in databases.

    But at that first revelation, I decided that in NO WAY any of this was going out on the internet for any reason. When homesteading there’s no need anyway – what would I do if I could see my woodstove from a cellphone (which I don’t bother owning)? Magically tell a more advanced robot than currently exists to put in another stick?

    1. Just as a curiosity, since the theme of different kinds of structutre has appeared : do you use a “serpentine” coil of copper pipe in that woodstove to help heat water ? That is another thing I have known to exist since my grandmother´s stove, and still get baffled when see people in the rural areas, which have a stove of that kind running almost all day long, complaining about having no hot water to wash/bath .

      The kind of simple, useful technologies that people seem to forget and suffer because of it.

      1. My old wood stove here in the UK doesn’t have a coil of copper pipe, it has a water jacket around the rear half (it’s cylindrical). All works by gravity, no pumps or electricity supply needed. Nothing’s for free of course, it means the stove takes longer and therefore more wood to get up to temperature.

      2. I’ve done that in the past as a source of hot water for baseboard heating. In a funny, I had an old one sitting on the porch which attracted law enforcement, as they thought it might be a still for moonshine. Hilarity ensued. I never used it for household hot water as the inevitable copper salts generated were a bit much for a lot of body contact. I’ve use various other methods for wash water, from propane to a couple of different electric approaches. Once you have enough solar energy…a lot of things aren’t so difficult to solve as they once were.

        If you do a good enough flue design, there’s actually not that much left over heat to extract without ruining the draft…It’s largely a matter of how you distribute the heat of the fire, and a liquid transfer medium can help in some cases for sure. I don’t use it anymore here – it becomes a maintenance problem I don’t need.

  6. >Monitoring for frequent baths that don’t fill the tub all the way would be a good start.

    And you can tell this from watering the garden with a hose or running old tap water through a laser cutter, or just running the water for a bit to draw cold water in from the burred pipes instead of the copper in the house.

    There is so much FUD in these things it is not even funny. About the only things I can think of in this house that take a semi consistent amount of water are the toilets.

    I was chatting with one of the tech’s in our cities water department. The single biggest cause of water loss in the city is leaks on their side. The branch meters don’t align well with the sum of the individual meters. Also, when they disconnect a house, they make a big deal out of pulling the meter. This really gets me as if I were corrupt it would be easy to replace the meter with a plumbing union and turn the street key back on. Street keys are simple things to make. If they left the meter and seals in place at least they could compare the last reading on record with the current reading.

    1. Seems like you don’t understand the research. The point is that you CAN tell the difference between all those things if you have enough data and use machine learning to filter out the noise.

      Watering the garden won’t look the same as filling a bathtub because the flow from the hose/sprayer will be different than the tub faucet. Not to mention the fact that you’re probably not watering the garden at night before you to bed.

      1. I suspect the algorithms they’re using need to be trained which means you first have to feed it “known” data. And as I stated in my posts above, getting meter data by the second isn’t really feasible without physically accessing the meter with specialized hardware and software….nobody is “hacking” your smart meter and getting readings by the second…most likely they can’t get any data at all unless they lift the lid and visually look at the register.

    2. Hi, author here. The key to making the classification work well was to define a triplet of values for each flow event – average rate, total volume, and duration. The combination of those three things was enough to classify many events – for example, in your laser cutter example, flow rate would be lower and duration higher than for a tub fill, for example. It’s surely not perfect, but it does work for many events. I wasn’t trying to introduce uncertainty or doubt – I’m reporting that you can most assuredly do it. :-) I appreciate the interest.

  7. This is one reason I do not care for smart metering like this. I don’t want someone with some environmental cause stuck up their rear end knocking on my door asking why I’m running the heat as high as I am or why I’m using so much water.

    Maybe I’m developing film in my basement. Maybe I am caring for someone who is too anemic to handle a typical thermostat setting in the winter. Maybe I have a large fish tank. It’s none of their business. If I pay my bill dutifully every month, what difference does it make?

    This is where the smart city proponents are taking us. Do we really want to go there?

    1. ” If I pay my bill dutifully every month, what difference does it make? ”

      California water crisis. Mr Dutiful is watering lawn/car in the heat of the day. Why should anyone care, we’ll just make more water.

      1. Water use is ultimately an energy concern, not a water concern. As the saying goes in the western US: Water doesn’t flow downhill. It flows to money.

        Given appropriate cause, people will find a way to get the water they need. What’s wrong with letting markets decide how water will be allocated? Do note that when natural gas distribution was deregulated, prices and availability dropped significantly.

        1. Maybe a reason is that ecosystems and endangered species do not usually participate in and bid for scarce water resources in water markets? Markets do not effectively price in these externalities, in the same way unregulated energy markets fail to price in externalities such as air pollution.

    2. I think people with a concern for the environment would be just as worried by the ability of the surveillance state to determine you have another two people staying over around the time that the local community is organising about contamination of local ground water/air/food by #insert

      The smart city proponents are more likely to try and sell you funeral plans when your heater use drops off suddenly and there is one less person flushing the toilet.

  8. This is a lot like what Sense is doing with monitoring of electrical consumption. They attempt to calculate a signature for each device in the house. Sometimes it works, sometimes it doesn’t.

  9. Even barring fine-grained (1s) monitoring that the author presents, simply knowing gross water usage, say on an hourly basis, is enough to know if the owner is on vacation. That can be sniffed easily with an RTL-SDR. Luckily I think most thieves are unlikely to hobby with RTL-SDR.

  10. This highly depends on the smart meter type. Some smart meters do not allow tracking to this level to prevent this type of analysis. They aggregate to buckets of 15/30/60 minutes. They will not give such small increments. That can be use to determine if a person is home, but not track the individual appliance.

  11. If this really were an issue , install a large cold water domestic water tank with a large range fill float switch. This will filter out any useable data from the meter, you will only get fills when the tank limit low is triggered thus hiding all bowel movements etc.

  12. I am sure that HAD have covered similar thing before. A guy was making a system to check if his granny is ok without installing cameras in her house. RPi with some AI to look for anomalities in water usage. Is this the same guy?

  13. I found exactly the same when I started doing detailed energy monitoring. I could see when my girlfriend came home (even if I hadn’t seen her on the cam outside the front door), when she made coffee and when the fridge door had been left ajar.

    It’s scary indeed and big data algorithms are much better at reading into these things as we are.

  14. I had a bad flapper valve in one of the toilets. Took a few days to bubble to the top of the list. I got a paper letter from the city suggesting I check for leaks because of water use in all 24 hours of the day.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.