Determining Location By Listening For Airplanes

An audio spectrogram is shown in the middle of the image, with intensity plots shown both above and below the spectrogram. A vertical yellow line runs through all three graphs.

Beyond what people deliberately disclose, there’s a surprising amount of information hidden in everyday messages, especially photos and videos. Even when people try to hide the location or time at which something was recorded, small details make these surprisingly difficult to obscure. One of the more well-known techniques correlates mains-frequency hum with power grid frequency variations. In a similar vein, [colsto] recently developed a powerful technique for precisely geolocating videos using the Doppler curves generated by passing aircraft.

Since aircraft are often audible for miles in all directions, they’re particularly likely to be inadvertently picked up in a video. The aircraft may not be discernible to the human ear, but the sound it makes can still show up in a spectrogram. To turn this data into a location, the pattern of Doppler curves needs to be correlated to an aircraft’s flight path. This is easy enough using public ADS-B records, which provide a timestamped history of each aircraft’s location, altitude, speed, and heading.

Having more than one Doppler curve in the recording simplifies the location process significantly: it’s possible to simply search for points at which Doppler curves would have been heard with the correct interval between them. It’s also important to exclude locations at which more Doppler curves would have been heard than were actually recorded – without this filter, most locations are around airports.

This eliminates most candidate areas, but it still leaves too many to manually check. The shape of the Doppler curve also provides a clue, though. The height of the curve is determined by the airplane’s speed, and the length by its distance. From an aircraft’s flight path, it’s possible to reconstruct the Doppler curve which would have been heard at any point around its flight path. These reconstructions can then be matched against the real curve to find the exact location. In a test, even a single Doppler curve was enough to locate a video without the help of any other clues.

This is just the latest in a long series of unexpected ways in which we’ve seen secrets shared, whether the secret in question is the location of a military base or the position of radar emitters.

Thanks to [Andrew] for the tip!

32 thoughts on “Determining Location By Listening For Airplanes”

  1. The video reminded me of a few from the last years. Including voxel projection. It’s be a fun project to reproduce, and see if it also works for different places and regions. The only real question I have for this is, what is the probability a video has one has multiple propeller airplanes in it? It got to be very low probability, but then, if all other options result in nothing, this is a great option.

    1. It reminds me of that affair nearly a decade ago with the “He will not divide us” banner that Shia LaBeouf hid somewhere in the middle of the country with a webcam, and was similarly doxxed by people online carefully watching airplanes in the sky in the background and using ADS-B.

      This is a more insidious doxxing tool. You can easily not film the actual image of the airplanes or contrails, but the sound may filter in.

    1. Might have been useful back in the days of tracking down terrorists who were releasing vhs and audio tapes from caves. But now the intelligence agency just quietly insets itself in your supply chain and sells you walkie talkies with bombs inside.

    2. I’d like to think this was computationally expensive enough that they’d need to find you particularly interesting in order to deem it worthwhile to follow though with this sort of thing. For a one off perhaps, but a lot of this can be precalculated as it’s happening. No doubt they’ve got computing power and storage capacity we couldn’t even Imagine.
      Most people are boring enough, that being considered as a filter would make anyone trying to hide, by artificially creating audio for communications, stick out to a deterministic AI workflow. Maybe adding in doppler audio to purposefully fake a location or send an algorithm on a wild goose chase with multiple targets tying up processing time.

      1. I’m sure someone can figure out a way to hash or index these curves for faster matching. The curves can be calculated and stored in real time as the planes are transmitting their ADS.

    3. Based on the recent SmartTV always-on-microphone stories, everything. But being technically able to and actually listening to you specifically are two different things. Though I’m sure someone is pushing to record everything and put it in a datacenter for later even if analyzing it now is impractical.

      I’d be more concerned with work / school computers spying on you. There have already been a few instances of schools recording students outside of school hours.

      1. I’ve worked for Sii Poland between 2023-2025. After carefully examining firewall logs it turns out the laptop they provided me for WFH did some really shady data transmissions every day around 3 AM, even if shut down and disconnected from charger (no entry in Windows task scheduler btw). After that discovery I started to keep their laptop outside in a plastic box on a balcony. Suddenly 3 months later I got voluntarily resigned from my job for some very vague “misuse of company’s property”. How they detected that, I don’t know but my spidey senses were tingling and honestly I’m glad I’m not slaving away anymore for that fat frenchie so that he can buy more footballers for his club. Now I’m back living with my parents in Krynica Morska and honestly no regrets. I’d rather walk to the beach and watch the sea every day than stress myself over some failed yocto build that got my resource manager angry.

        1. Could be IME checking in to see if there are any maintenance tasks the laptop needs.

          Also, “suddenly” and “3 months later” don’t add up.

    4. What struck me was that apparently some camera high up on a pole on a pier in the middle of nowhere is recording sound.
      And making that sound available.
      Same as that many doorbell cameras seem to constantly record and stream-home audio. (and incidentally, can you be sure it does not hear what you say inside too? I would not advise assuming it can’t).
      And yes phones have mics too, but there’s more attention to them, and a possibility to check.
      But of course cars have internal mics too, and televisions, and remotes, and people pay lots of money to put orbs with mics all over their houses.

      And I gather that in lots of countries public transport that has ton of cameras also have microphones, so watch what you say because now AI can analyse and highlight certain topics so they don’t even have to ‘just happen to be listening’, no, you have to assume all is not only monitored but also judged and evaluated.

      I wonder if they secretly invented something yet to listen to you from space, through reflection or RADAR or something. But anyway, soon with every phone having sat connections they can hack phones (or use supplied backdoors) and listen in everywhere anyway I guess.

    1. Yeah I’ve never watched this guy before but selling yourself as OSINT, making a tool OSINT people would love, and then releasing 0% of it was pretty annoying. It’s weird, not even like he is trying to sell it. He seems to just want to brag…

      1. Perhaps he’s excited to share what he’s gotten to work, but the source is still so messy (in his eyes, not objectively) that it’d be embarrassing to put out in the public eye.

        1. Or maybe, he realized, that making stuff Open Source/free often brings out the “you OWE me, because I wanna use your shit” crowd.. and don’t wanna bother with questions etc.

          1. The FOSS koan:

            Please return the unused portion of the code, and we’ll return the unused portion of the money you paid for it.

    1. To compare that to this is a stretch that exceeds the width of the known universe, thereby showing we are wrong about the size of it and the Big Bang postulation.

      Quite a breakthrough.

  2. cool but i think we learned this in the movie The Fugitive with Harrison Ford, when Tommy Lee Jones character asks something like, let me hear the sound of a L train. One project I have in my personal dta haven’t shared anywhere yet, is to put a outside mic on all four corners of the house and see what can be heard. planes, trucks, cars. This will be handy.

  3. Upon reading the title, my weird brain expected super slow, terrible version of GPS using an ADS-B receiver and a locally stored, refreshed-as-frequently-as-practical flight schedule.

    Likely a better fit for halfbakery than hackaday.

  4. I watched this video a couple days ago and was so thrilled as he blew my mind over and over again with the elegance of this analysis and the way that the factors in play were osint accessable. It weirded me out because of the realization that every device with a microphone (I’m looking at you TVs) could be accoustically located/dead-reconed to a area so small it made me uncomfortable.

    But then as I drifted off later, or maybe this morning as I recall it… It would be possible to do the math for accoustic sigs as a grid for every hexagon in the US and then make a lookup table for accoustic sigs, and then its just hit an index for estimates and calc the details when needed. Seems like something feasable for any agency or for a well-computed hobbyist.

  5. I was impressed by the video. Seems like a lot of work that involved human judgement and guidance. But yeah, if an intelligence agency wanted to geolocate simply an audio recording, sounds like they could. I wonder how many job offers this guy got from spooky agencies. Or is that not how this sort of recruiting works.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

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