aural cryptography

David Honig dahonig at
Tue Apr 8 13:28:18 EDT 2003

At 11:50 AM 4/8/03 -0400, John Kelsey wrote:
>I was re-reading the original visual cryptography paper last night, and had 
>an odd thought: Why couldn't we do something similar with sounds?  The 
>human ear/brain is pretty good at pulling patterns out of noise; would it 
>be possible to randomly embed half of a low-quality voice channel in each 
>of two sound channels, so that they didn't sound obviously bad apart, but 
>when played at the same time, would allow the listener to hear a spoken 
>message pretty clearly?
>It seems like you could even do some pretty weird things with this, like 
>embedding the signal in four or five sound channels, or embedding them in 
>such a way that the speakers on the different channels had to be a certain 
>distance apart for the embedding to work.
>So my questions are:
>a.  Is this really possible?  Or am I missing something?
>b.  Has this been done in the open literature?  (It seems like the sort of 
>thing that would have been really useful for, say, radio broadcasts that 
>were intended to be received by spies.)

You could put nominally subthreshold speech in each of two
aural channels and let the brain sum them.  But in practice,
no.  First, individual thresholds vary.  Second, voice
is not a constant-amplitude signal.  

Also of course, as with visual perception, 
a machine can do this.  Indeed, for audio, 
spurious nonlinear artifacts might do it

The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to majordomo at

More information about the cryptography mailing list