EarSpy can eavesdrop on your phone conversations using motion sensors
Mobile security is kind of like a highway: new potholes form every day and its throughput capabilities are highly dependent on the drivers taking care not to cause a pile-up. Whether these crashes are caused byresearchers sniffing out a new vulnerability,players down the security chain not doing their part, or worse, all of it’s happening day in, day out. A group of researchers from some of America’s most reputed academic institutions have now developed an attack named EarSpy, designed to capture what users say through curiously crafty means.
This effort is being carried out jointly by experts at the University of Dayton, New Jersey Institute of Technology, Rutgers University, Texas A&M University, and Temple University. Researchers have attempted to gather vibrations from a phone’s loudspeaker in the past, but this particular attack is effective even when the user is holding the phone to their ear, SecurityWeekreports.
The research team tested out its EarSpy exploit by playing a series of voice samples through the OnePlus 7T and the OnePlus 9 and analyzed accelerometer data with several neural network tools and recorded astonishingly accurate results with regards to identifying the gender of the speaker as well as the words spoken. By contrast, the data was hard to capture on older OnePlus models due to the lack of a stereo speaker system, the researcherssaid in their paper.
Newer Android versionshave a more robust security apparatus, making it exceedingly difficult for malware to get the requisite permissions. But EarSpy attacks can still bypass these built-in safeguards as raw data from a phone’s motion sensors are easily accessible. Although more manufacturers are now placing limits on obtaining data from the device’s sensors, EarSpy researchers believe it’s still possible to infiltrate the device and eavesdrop on a conversation.
As for the effectiveness of this attack, the researchers say EarSpy could correctly tell the difference between males and females in 98.66% of samples in one particular set played through the OnePlus 7T. Across both phones, two sample sets, and multiple analysis models, gender recognition is fairly accurate with the lowest reading being 65.53%. Furthermore, EarSpy could detect the person’s identity with a ridiculous 91.24% top accuracy rate — nearly three times better than a random guess.
When it comes to actually understanding what was spoken, though, accuracy falls significantly. Recognition models were tested against samples featuring actors reciting a sequence of digits — not out of place if we’re talking about sharing personal identifiable information like a credit card number over the phone. The best performer achieved a 56% hit rate, though researchers say this is still five times more accurate than making a random guess.
The paper’s authors say that while the actual impact of speakerphone vibrations on raw accelerometer data is relatively low and that algorithmic word detection off that data is extremely spotty, adversaries who use the EarSpy exploits can still determine the key components of what is being spoken and who is speaking it.
In theory, EarSpy could be leveraged by malware that has infiltrated the device and relay accelerometer data back to the source of the attack. This report highlights the importance of additional hardware safeguards, especially with components like motion sensors that may not seem like easy targets at first glance.
To remedy this potential vulnerability inmodern-day smartphones, the researchers recommend smartphone makers to position motion sensors away from any source of vibrations while also reducing sound pressure during phone calls.
UPDATE: 2023/01/02 15:30 EST BY JULES WANG
Clarifications
This article has been updated to provide more specifics about the data we’ve cited from the paper.
Pixel 10 Pro XL charges faster wirelessly
From faster storage to better speakers
Carriers get the upper hand
Google’s made several improvements over the years
Broader branding hints at wider paid-tier ambitions
The note-taking app I should have used all along
