I'm Micah Loewinger filling in for Brooke Gladstone. Okay, so this next mystery is a little different. It has to do with an obscure form of audio forensics, a technology called Electrical Network Frequency analysis, or ENF.
Jen Munson: Hi.
Micah Loewinger: That's Jen Munson. She's On the Media's technical director. Her job is to make the hosts, producers, reporters, and the people we speak to sound as clean and clear as possible.
Jen Munson: My approach is mostly to find the thing that I like in someone's voice and bring that out.
Micah Loewinger: I called her up to tell her about ENF analysis, though she didn't know that at the time. I just said I was working on an episode about audio mysteries.
Jen Munson: Audio mysteries.
Micah Loewinger: [laughs] I told her that I sent a scientist some recordings of me interviewing people on our show. Just my side of the conversation, just my voice. Using ENF analysis, this researcher was able to tell me the day and time, almost to the exact second that I recorded each interview.
Jen Munson: Really?
Micah Loewinger: Yes.
Jen Munson: That's in the metadata, because you sent him a digital file.
Micah Loewinger: I rebounced it so that if you checked the metadata of the file, it would be when I made that file, not when I recorded it.
Jen Munson: Interesting. Forensic audio is really fascinating to me. I would think there would be a way to compare it to known other recordings, traffic sounds, like environmental sounds around you.
Micah Loewinger: That's a good guess, but that's not right though.
Jen Munson: That's not right?
Micah Loewinger: No.
Jen Munson: Tell me, now I want to know.
Nasir Memon: Hi, I'm Nasir Memon. I'm a professor here at NYU's Tandon School of Engineering in the Computer Science and engineering department.
Micah Loewinger: Nasir oversees the group at New York University that published papers on ENF analysis as recently as this year. They get funding from the Defense Advanced Research Projects Agency, AKA DARPA, the research arm of the Department of Defense. This type of audio forensics has been studied in academia for a couple decades now, but its used by law enforcement is what caught his eye.
Nasir Memon: To my understanding, the first folks that had done it was the London Metropolitan Police.
Micah Loewinger: In 2010, a specialist at the Metropolitan Police Department described ENF as, "The most significant development in the field since techniques were developed to analyze the Watergate tapes." Nasir explained to me that for ENF analysis to work, he needs to find something specific in the recordings I sent him, a bit of interference, that Jen is very familiar with.
Jen Munson: You'll see on the graph, you're getting just the low rumble.
Micah Loewinger: I had Jen use a fancy audio equalizer tool to look at the different frequencies in a recording I sent to Nasir's team.
Jen Munson: There's a little bit of this hum 60 hertz. Can you hear that?
Micah Loewinger: I think it's imperceptible in this recording.
Jen Munson: My ears are tuned to hear it.
Micah Loewinger: Audio engineers will tell you that this 60 hertz hum contaminates all kinds of recordings.
Jen Munson: That's the first thing I'm approaching, is getting rid of that sound. You hear it a lot on guitar amps. You hear it with--
Nasir Memon: My refrigerator, a TV--
Jen Munson: Elevators, in buildings, you hear it.
Nasir Memon: Everything is interconnected, right?
Jen Munson: It's very common.
Micah Loewinger: Many of our electrical things all around us are constantly buzzing at a 60 hertz or a harmonic, like 120 hertz, 180 and so on. What we're hearing or not hearing is the electrical grid. The companies that manage our power, in my case, Con Edison in New York, are required by law to maintain that 60 hertz output.
Nasir Memon: It's unable to keep it exactly at 60, because the consumption is varying. Lights turn on and off and people turn on their devices. It's trying to maintain and cater to the load. It doesn't want to produce too much electricity.
Micah Loewinger: The demand for electricity is constantly changing based on what's plugged in.
Masir Memon: The production mechanism is trying to keep pace with it, and it doesn't succeed in maintaining it to exactly 60. It becomes 59.8, 60.1.
Micah Loewinger: If you were to map the frequency over time, it would not be the straight 60 hertz. It would be this ever so slightly wiggle.
Nasir Memon: Yes, ever so slightly wiggle. The utility companies, they have to measure this and report it to the government, but we can measure it too.
Micah Loewinger: Nasir's former grad student, [unintelligible 00:26:08] built a very simple computer that records the wiggle from the grid every second or so. When I sent [unintelligible 00:26:15] my recordings, he isolated the 60-ish hertz hum, which might have come from my laptop charger, plugged in a few feet from my microphone.
Nasir Memon: He would pull out the data for the last three months that we've been capturing.
Micah Loewinger: I told [unintelligible 00:26:30] that I often recorded my interviews around a couple weeks before they aired, which a forensic specialist would determine pretty quickly anyway, but I told him that mostly so he wouldn't have to spend unnecessary time cycling through years of data.
Nasir Memon: Then he would run an algorithm on the sound using a sliding window. Every 20 seconds, he slides it over, and at some point, matches, matches, matches.
Micah Loewinger: Which is how he guessed the time of three of my recordings within around 10 seconds each.
Jen Munson: Wow.
Micah Loewinger: It's such a one man's trash is another man's treasure thing, that for you and audio engineers, the thing that makes your job slightly harder is actually this forensic fossil that can be dug up to glean information about when something happened, when the recording was made.
Jen Munson: I'm blown away.
Micah Loewinger: Even though ENF has been around for nearly two decades, it doesn't seem to have caught on in any significant way in the US. I reached out to an editor at Bellingcat, the cutting edge investigative outlet known for its use of data and tech. They knew about ENF analysis, but weren't familiar with journalists using it. I also couldn't find any court records mentioning its use by American law enforcement. Catalent Gregor, one of the early developers of ENF told me it's often used for checking to see if media has been edited or tampered with. You can compare the hum in a piece of media to the data from the grid to see if the audio has been spliced or rearranged.
Other scholars have referenced the Osama bin Laden cave videos as a hypothetical application. Investigations in which the ability to learn which electrical grid a person is near might offer up new leads, but the issue here is that the technology only works if, A, the hum is captured in the recording, which is a bit of a crapshoot, and B, you have access to the right grid data. Nasir Memon.
Nasir Memon: If I was running this in an intelligence agency, I would make sure I'm capturing everywhere in the world.
Micah Loewinger: Do you think the American intelligence agencies are interested in ENF?
Nasir Memon: Intelligence won't tell me. Even if I knew, I may not be able to tell you as well, which I don't.
Micah Loewinger: Do I believe you?
Nasir Memon: [laughs] No, I don't have any secret clearance, anything of that.
Micah Loewinger: You mentioned that the study was funded by DARPA, and I understand that similar research has been done completely independent of DARPA, but it does seem like you are helping develop a technology that could be used for surveillance.
Nasir Memon: Right. We are scientists. We like to further science. Science can be used for good and bad. We just leave that question aside quite often, because it just gets very, very, very complicated. The answers are not clear.
Micah Loewinger: He told me, at one point in his team's research, he had considered collecting way more data.
Nasir Memon: Even different countries, make it public, put a tool whereby you submit a video and then I'll tell you what time it was taken. Then I thought, that's going too far. That's going too far because the ethical issue started coming up then to me.
Micah Loewinger: What if a stalker wants to try to track somebody down using videos you posted on Instagram?
Nasir Memon: That's why I didn't do it. I'm not trying to say I'm a very ethical person. If there was money there in it, maybe I would've done it. I don't know. There was no reason to do it. For what purpose? I did not.
Micah Loewinger: I have gotten you to admit that you're corruptible.
Nasir Memon: [laughs] Well, we all can be. Let's put it that way.
Micah Loewinger: I don't really know how useful ENF analysis is for mass surveillance. There are just far better ways to track people, like GPS or the type of stuff you can legally buy from a data broker. Even little visual clues in the back of a selfie can lead a dedicated sleuth to figure out where you live. What drew me to learn more about ENF is the poetry of it. Think about it, every time you turn on a light, or plug in your phone, or vacuum your rug, or blow dry your hair, you're contributing to that ever so slightly wiggle as the grid adjusts itself to our needs. It's a barely audible symphony that we're all playing a part in.
The American electrical grid, which has been called the largest machine in the world, is a pulsating map that should remind us of just how interconnected we all are. Coming up, our final mystery is about another map made of sound. That no human, not even Jen Munson, can hear. This is On the Media.
