Cold War spies planted bedbugs in walls, lamps and telephones. Now, scientists warn that the cables themselves can listen. A fiber-optic technique used to detect earthquakes can record the faint vibrations of a speech, researchers reported this week at the general assembly of the European Geosciences Union.
Freely available artificial intelligence (AI) software transformed fiber optic data into understandable real-time recordings. “Not many people know that fiber optic cables can detect acoustic waves,” says Jack Lee Smith, a geophysicist at the University of Edinburgh, who presented the research.
"We show that in almost every case where you use these fibers, this could be a privacy issue."
Optical fibers can record sound through a technique called distributed acoustic sensing (DAS). Using a machine called an interrogator, researchers shoot laser pulses down a cable and record the pattern of reflections that return from tiny glass defects along the optical fiber.
When an earthquake's seismic wave passes through a section of the optical fiber, it stretches and compresses these defects, leading to shifts in the reflected light that researchers can use to create an image of an earthquake.
DAS essentially turns a fiber-optic cable into a long chain of seismometers that can detect not only earthquakes, but also the noises of volcanoes, cars, and more. And although scientists have created special fiber-optic lines specifically for research, DAS can also be performed on “dark fibers” – unused strands in the fiber-optic web that crisscrosses cities and oceans, carrying the world’s internet traffic.
DAS can also be used for monitoring, as Smith and colleagues show in their study. They conducted a field test using an existing DAS setup used to study coastal erosion. They placed a speaker next to the cable and played clear tones, music, and speech.
Human speech contains frequencies ranging from a few hundred to several thousand hertz. The low end of the range could be extracted from the data “even without any preprocessing,” Smith says.
“You can easily see the acoustic waves.” Capturing higher-frequency speech requires some post-processing, but it is also possible.
Directly transferring the data to Whisper, a free AI transcription tool, provided accurate, real-time transcription. However, this technique only worked for cables that were exposed on the surface, up to 5 meters from the speaker. Burying the cable under just 20 centimeters of soil is enough to muffle the speech.
doi: 10.1126/science.z8iynq5
Although the press releases will range from very select to rare, I said I'd pass...because sometimes the editors hide.
