[HN Gopher] EarSpy: Spying on Phone Calls via Ear Speaker Vibrat...
       ___________________________________________________________________
        
       EarSpy: Spying on Phone Calls via Ear Speaker Vibrations
        
       Author : simonpure
       Score  : 59 points
       Date   : 2022-12-28 14:27 UTC (2 days ago)
        
 (HTM) web link (www.securityweek.com)
 (TXT) w3m dump (www.securityweek.com)
        
       | anyfactor wrote:
       | This reminds of how Helen Keller learned to communicate. The
       | technique is called Tadoma[0].
       | 
       | > Tadoma is a method of communication used by deafblind
       | individuals, in which the deafblind person places their little
       | finger on the speaker's lips and their fingers along the jawline.
       | The middle three fingers often fall along the speaker's cheeks
       | with the little finger picking up the vibrations of the speaker's
       | throat.
       | 
       | [0] https://en.wikipedia.org/wiki/Tadoma
        
       | 95014_refugee wrote:
       | This is (very) old news, and responsible vendors have had
       | mitigations in place for a long time now.
        
       | hackernewds wrote:
       | > In the gender recognition test, whose goal is to determine
       | whether the target is male or female, the EarSpy attack had a 98%
       | accuracy. The accuracy was nearly as high, at 92%, for detecting
       | the speaker's identity.
       | 
       | When it comes to actual speech, the accuracy was up to 56% for
       | capturing digits spoken in a phone call.
       | 
       | "[This] accuracy still exhibits five times greater accuracy than
       | a random guess, which implies that vibration due to the ear
       | speaker induced a reasonable amount of distinguishable impact on
       | accelerometer data," the researchers said.
       | 
       | this seems not to have any practical significance. and
       | recognizing numbers (only 56%) of the time is practically still
       | as good as guessing, and not useful, in the context of spying,
       | without the spoken words
        
         | dannyw wrote:
         | Some OTP codes can be sent via phone calls, and keep in mind
         | this is just a research prototype. It's certainly plausible
         | that improvements can result in higher accuracy.
        
         | messe wrote:
         | > this seems not to have any practical significance. and
         | recognizing numbers (only 56%) of the time is practically still
         | as good as guessing, and not useful, in the context of spying,
         | without the spoken words
         | 
         | If that accuracy is at the single digit level, then it's
         | significantly better than guessing, which would yield a 10%
         | accuracy. 56% would be good enough to capture a 4-digit OTP
         | code more than 1/8 of the time.
        
         | codeddesign wrote:
         | "whether the target is male or female, the EarSpy attack had a
         | 98% accuracy"
         | 
         | I guess the other 2% are "gender neutral".
        
       | anfractuosity wrote:
       | I came across this paper recently 'mmSpy: Spying Phone Calls
       | using mmWave Radars' which says 'mmSpy performs sensing of
       | earpiece vibrations using an off-the-shelf radar device that
       | operates in the mmWave spectrum (77 GHz, and 60 GHz).'
       | 
       | There's a video with some audio samples here -
       | https://www.youtube.com/watch?v=EuuXAfIYo04
       | 
       | The Ti dev boards are around ~PS200 if I recall correctly.
        
         | 0ct4via wrote:
         | Link to the paper, since you didn't provide one...
         | 
         | https://www.cse.psu.edu/~mkg31/papers/mmSpy.pdf
        
       | jonpalmisc wrote:
       | If you find this interesting, you may also like some of Mordechai
       | Guri's work: https://cyber.bgu.ac.il/advanced-cyber/airgap
        
       | dessant wrote:
       | It's worrying that any website can access real-time data from
       | sensors by default in Android browsers, and there is no option to
       | deny access to these APIs in Firefox for Android and Samsung
       | Internet.
       | 
       | https://sensor-js.xyz/demo.html
        
         | commoner wrote:
         | Firefox's fingerprinting protection disables these sensors and
         | shows no data in the demo, even on Android. To enable this
         | feature, go to about:config and set
         | privacy.resistFingerprinting to true:
         | 
         | https://support.mozilla.org/en-US/kb/firefox-protection-agai...
         | 
         | This feature is enabled by default on some Firefox forks like
         | Mull.[1] On Android only, the release channel of Firefox
         | disables about:config entirely,[2] but Firefox Beta and Nightly
         | have about:config available.
         | 
         | [1] Mull: https://f-droid.org/en/packages/us.spotco.fennec_dos/
         | 
         | [2] Please vote to remove this ridiculous restriction at
         | https://connect.mozilla.org/t5/ideas/firefox-for-android-abo...
        
         | berkes wrote:
         | Looking at this page makes me think that it's probably possible
         | to detect _that_ (not what) people are speaking or typing.
         | 
         | I placed my phone on the surface. And did see the numbers jank
         | when I started typing. Then I left it at rest and said a loud
         | "Booo!" and saw the numbers jump again. To confirm, I'd need to
         | plot the numbers and do some statistical analysis, which is far
         | too much work for me now :D
        
       | lfodofod wrote:
       | This attack is nothing new, has been endlessly discussed by the
       | grapheneos team. An accelerometer is just a shitty microphone,
       | even a camera can be a shitty microphone.
        
         | 0_____0 wrote:
         | I have actually used a (nice, high fidelity and sample rate)
         | accelerometer as a mic. Rolloff is somewhere around 400Hz. Just
         | popped the accel data into audacity and played it back.
        
       | karbjlnms wrote:
       | Didn't Android already restrict access to accelerometer data by
       | default?
        
       ___________________________________________________________________
       (page generated 2022-12-30 23:01 UTC)