Work business call

(© Drobot Dean – stock.adobe.com)

In a nutshell

  • Your microphone is leaking conversations: Digital microphones in laptops, phones, and smart speakers unintentionally broadcast electromagnetic signals that can be intercepted up to 2 meters away, even through walls.
  • The attack is surprisingly accessible: Researchers achieved over 94% accuracy in speech recognition using simple equipment like copper tape antennas, making this vulnerability exploitable by anyone with basic technical knowledge.
  • Your “off” microphone might still be listening: Testing revealed that microphones often activate automatically when playing audio or video content, and some remain active even when apps appear muted.

GAINESVILLE, Fla. — That private work call you had yesterday? Someone next door could have been listening to every word — not through thin walls, but by intercepting invisible signals leaking from your laptop’s microphone. New research reveals that millions of everyday devices are inadvertently broadcasting conversations through electromagnetic radiation that attackers can capture and decode from up to two meters away, even through concrete walls.

Digital microphones in laptops, smartphones, and smart speakers are unintentionally acting like tiny radio transmitters, allowing eavesdroppers to monitor conversations with frightening accuracy. Researchers at the University of Florida achieved up to 94.2% accuracy in recognizing spoken digits through a 25-centimeter concrete wall, with some transcriptions having error rates as low as 6.5%. Even more troubling: this surveillance can happen when users believe their microphones are turned off.

How Modern Microphones Became Accidental Transmitters

Nearly every electronic device today uses PDM (pulse-density modulation) digital microphones, which are tiny components just a few millimeters wide that have replaced older analog versions. Unlike traditional microphones that convert sound into varying electrical signals, these digital versions have built-in mini-computers that transform audio directly into patterns of digital pulses.

Louder sounds create more frequent pulses, while quieter sounds generate fewer pulses. The problem? These pulse patterns create electromagnetic radiation at multiple frequencies, essentially turning every microphone cable into an unintentional antenna broadcasting your conversations.

The research team, led by Sara Rampazzi, Ph.D., discovered that standard radio equipment can pick up these signals and decode them back into recognizable speech. As the researchers explain: “each harmonic of these digital pulses retains acoustic information, allowing the original audio to be retrieved through simple FM demodulation using standard radio receivers.”

silver MacBook Pro on white table
An alarming new study shows how your laptop microphone could be secretly broadcasting your conversations. y(Photo by Christopher Gower on Unsplash)

Real-World Testing Reveals Widespread Vulnerability

Researchers tested five major microphone brands commonly found in consumer electronics, including products from Knowles, STMicroelectronics, and TDK-InvenSense. All showed the same vulnerability, with digit recognition accuracy above 98% and speaker identification above 97% across different devices.

Their experiments extended beyond laboratory conditions to real-world scenarios. In one test, they placed an attacking antenna in an adjacent room behind a 15-centimeter-thick plasterboard wall, a typical office building setup. Even through this barrier, they could classify spoken digits with 95.5% accuracy and achieve transcription error rates as low as 6.5%.

The team also uncovered a disturbing reality about microphone activation. Testing popular platforms like Spotify, YouTube, and Amazon Music revealed that microphones often turn on automatically when audio or video content plays, regardless of user settings. Some remained active even when services appeared muted, creating ongoing surveillance opportunities.

Simple Tools Enable Sophisticated Spying

One of the study’s most alarming discoveries involves how inexpensive and accessible these attacks can be. Researchers successfully captured conversations using improvised antennas made from copper foil tape connected to basic amplifiers. Even with this crude setup, they achieved digit recognition accuracy above 94% and could transcribe speech well enough to understand most conversations.

“With an FM radio receiver and a copper antenna, you can eavesdrop on these microphones. That’s how easy this can be,” said Rampazzi, a professor of computer and information science and engineering at UF, in a statement. “It costs maybe a hundred dollars, or even less.”

Unlike sophisticated cyber attacks requiring advanced technical skills, this vulnerability can be exploited with equipment available at electronics stores. The accessibility makes it particularly concerning for corporate espionage, stalking, or criminal surveillance activities.

Current defense strategies prove largely ineffective. Common protective measures like signal filtering and noise reduction actually made the electromagnetic signals easier to intercept in some cases. Only a hardware-level solution called spread-spectrum clocking showed promise. When researchers tested this method with just 1% timing variation, they reduced attack effectiveness from over 70% to less than 5%. However, implementing this fix would require manufacturers to redesign their products entirely.

A demonstration of the microphone attack through a 15 cm wall. (Credit: Sara Rampazzi)

Privacy in an Age of Accidental Surveillance

The vulnerability transforms trusted devices — laptops used for work calls, smart speakers in our homes, phones we carry everywhere — into potential surveillance tools that can be exploited by anyone with modest technical knowledge and basic equipment. From confidential business meetings to intimate family conversations, virtually any discussion near an electronic device could be intercepted remotely.

Corporate and national security implications extend far beyond individual privacy concerns. Sensitive discussions in offices, boardrooms, and government facilities could be monitored without any physical intrusion or malicious software installation. Unlike traditional cybersecurity threats targeting network connections, this attack exploits the fundamental physics of how modern microphones operate.

The researchers responsibly disclosed their findings to affected manufacturers. While some companies acknowledged the issue, one vendor stated: “Our products are designed to provide reasonable and industry-standard protection against harmful interference and to comply with regulations for electromagnetic compatibility and radio frequency emissions set by the US FCC and other governments around the world. The present design is aligned with industry best-practices and customer requirements.”

As our homes and workplaces fill with an ever-growing array of connected devices, the revelation that microphones may be inadvertently broadcasting our most private moments represents another erosion of personal privacy. In an era where digital surveillance already permeates daily life, discovering that our devices are physically leaking conversations adds a new dimension to privacy concerns that extends well beyond the digital realm.


Paper Summary

Methodology

Researchers tested electromagnetic side-channel attacks on five different PDM microphone brands using various antenna configurations, from professional equipment to improvised copper tape antennas. They conducted experiments at distances up to 2 meters, testing through different wall materials including plasterboard and concrete. The team used multiple evaluation methods including machine learning models for speech recognition, digit classification, and speaker identification, as well as commercial speech-to-text services from Microsoft and OpenAI.

Results

The study achieved up to 94.2% accuracy in recognizing spoken digits through a 25cm concrete wall, with transcription error rates as low as 6.5% in optimal conditions. All five tested microphone brands showed vulnerability, with digit classification accuracy above 98% and speaker identification above 97% across different devices. The attack worked effectively even with cheap improvised antennas made from copper tape, achieving over 94% accuracy in speech recognition tasks.

Limitations

The attack’s effectiveness decreases significantly with distance, dropping to around 10% accuracy beyond 4 meters. Signal quality varies greatly depending on device design, particularly the length of internal microphone cables—devices with shorter cables (like some headsets) showed much lower vulnerability. The method requires some technical knowledge to identify optimal frequencies and may be affected by electromagnetic interference in certain environments.

Funding and Disclosures

This research was funded by JSPS KAKENHI Grant Number 22H00519, JST CREST JPMJCR23M4, and a gift from Meta. The researchers followed responsible disclosure practices, contacting all affected device manufacturers. Three vendors responded, with two assessing the vulnerability and one stating their products meet current industry standards.

Publication Information

The research paper, “Sound of Interference: Electromagnetic Eavesdropping Attack on Digital Microphones Using Pulse Density Modulation” by Arifu Onishi, S. Hrushikesh Bhupathiraju, Rishikesh Bhatt, Sara Rampazzi, and Takeshi Sugawara. Research artifacts and demonstration materials are available at https://cpseclab.github.io/Soundofinterference/ with datasets available at https://zenodo.org/records/14736347.

About StudyFinds Analysis

Called "brilliant," "fantastic," and "spot on" by scientists and researchers, our acclaimed StudyFinds Analysis articles are created using an exclusive AI-based model with complete human oversight by the StudyFinds Editorial Team. For these articles, we use an unparalleled LLM process across multiple systems to analyze entire journal papers, extract data, and create accurate, accessible content. Our writing and editing team proofreads and polishes each and every article before publishing. With recent studies showing that artificial intelligence can interpret scientific research as well as (or even better) than field experts and specialists, StudyFinds was among the earliest to adopt and test this technology before approving its widespread use on our site. We stand by our practice and continuously update our processes to ensure the very highest level of accuracy. Read our AI Policy (link below) for more information.

Our Editorial Process

StudyFinds publishes digestible, agenda-free, transparent research summaries that are intended to inform the reader as well as stir civil, educated debate. We do not agree nor disagree with any of the studies we post, rather, we encourage our readers to debate the veracity of the findings themselves. All articles published on StudyFinds are vetted by our editors prior to publication and include links back to the source or corresponding journal article, if possible.

Our Editorial Team

Steve Fink

Editor-in-Chief

John Anderer

Associate Editor

Leave a Reply