Schneier on Security
A blog covering security and security technology.
« Document Verification |
| VOIP Encryption »
April 5, 2006
Security Applications of Time-Reversed Acoustics
I simply don't have the science to evaluate this claim:
Since conventional sound waves disperse when traveling through a medium, the possibility of focusing sound waves could have applications in several areas. In cryptography, for example, when sending a secret message, the sender could ensure that only one location would receive the message. Interceptors at other locations would only pick up noise due to unfocused waves. Other potential uses include antisubmarine warfare and underwater communications that benefit from targeted signaling.
Posted on April 5, 2006 at 1:06 PM
• 36 Comments
To receive these entries once a month by e-mail, sign up for the Crypto-Gram Newsletter.
There have been published crypto papers on aural secret sharing, which works in a way analogous to visual secret sharing. It looks to me like this might be similar.
Doesn't help that the article is either poorly translated from French or written by a non-fluent speaker. From what I can tell, though, it's nothing more cryptographically powerful than a whisper chamber.
Would that be "the cone of silence"?
Why does this make me think of cold fusion?
Boston's museum of science has a set of ultra-sonic speakers that focus ultrasound in such a way that it makes air produce audible sound in a specific spot. If you stay exactly in the focus point of this thing you hear music, if you shift a foot away in any direction, you don't hear anything. Impressive. I can imagine something like that being used to deafen everybody on a submarine, but what does it have to do with cryptography?
"Why does this make me think of cold fusion?"
It doesn't sound like cold fusion to me. Apparently, "time reversal" is established terminology in nonlinear optics. The technique sounds to me more like an acoustic analog of using a laser beam to transmit information. I'm not sure if scattering would allow a non-intended recipient to pick up a weak version of the signal.
@max: As far as I understand this issue, it doesn't have much to do with cryptography, that is with securing data against being read by unauthorized person who happen to get a hold of it.
To me it sounds much more like simulating a physical connection like a cable. And it should be vulnerable to a man-in-the-middle approach, so classic cryptography is still needed. (Mitm doesn't work against submarine-deafening, of course ;)
IIRC, the 'focussed ultrasound' in the museum in question is an application of "beating" - two waves that coincide will "beat" with a frequency equal to the difference of the frequencies of the two waves. Ultrasound can be effectively focussed into a 'beam', and so the beating effect occurs in the crossover of the two beams, which can be made as narrow an area as you might like.
Unless I've misunderstood completely what I've read on the subject, good cryptography relies on the fact that the attacker could learn *all* the details about the algorithm and monitor *all* data transmitted and *still* not be able to decrypt the data in a reasonable time frame so long as he lacks the key. The *only* secrets involved should be the private keys and the plaintext message.
I don't see where this nifty focusing trick adds any specific value to cryptography. It seems like snakeoil designed to deceive the user into thinking his data is safe from interception. As soon as the attacker learns where to stand, this little trick is useless.
'Would that be "the cone of silence"?'
Perhaps....or perhaps instead it is "Hover Cover"...
It sounds like they have a technology for a system of line of sight transmission - it would make listening in harder, but isn't cryptography.
I would think that, if this were worth doing, someone would already be doing it with IR lasers. (They probably are.)
This is what happens when people stick a microphone in a scientist's face and say, "What is it good for?" The guy who invented the laser came up with very few of the uses we put them to and a lot of ideas that never came about.
When I read "time-reversed accoustics," I thought this was something about playing records backwards for their secret messages. This encryption method appears to have been broken more times than it has been employed, however:
Maybe speakers that only direct sound to a single area and not a whole room?
Like headphones that you dont have to wear on your head.
This is an attempt to go from 'line-of-sight' to 'point-of-sight'. It's a good idea, but not one you should rely on--it's just a neat steganographic trick.
Canceling sound ultimately requires overcoming the time delay back to a sound's source and predicting what a sound look like as it is created at that source. Time-reversing simply allows canceling predictable sound in the reverse direction (back to the source). You don't need time-reversing for cancelling sound in the forward direction--only prediction. This latter case applies to noise canceling headphones (since they are not cancelling sound back to the source). Still, they have to predict what the sound will look like in the forward direction, and that is why they only tend to work on low frequencies (which are regular and allow plenty of time for prediction). White noise and spurious higher frequencies aren't electively canceled. Its in the the frequent variations of the higher frequencies in which most our information is contained.
So what can be done to predict future data,and how can that apply to security? Don't we make our secrets irregular so that common phrases won't make a cypher easy to crack? It seems to be that the patterns of data have already been widely studied. Compare the informational entropy of sound waves (a cyclic pattern) to that of textual data. Compare the compression methods. If no one has proposed using aspects of audio compression in cryptology, then there probably isn't much here.
I concurr with MSB; if it works perfectly in real-world conditions, it's analogous to using a laser to send your message instead of a lightbulb. Unlike a laser, you can't transmit individual quantum mechanical packets of information.
The acoustic science behind this article has to do with the coherency of the sound waves after passing through a disordered medium vs. passing through an ordered medium. Coherency would allow the sound to be directed to a point, or at least along a narrow path to its intended target.
Another focusing technology is hypersonic waves, which encapsulates different wave streams inside other streams. This Time-Reversal Acoustics seems to be simpler and requires less energy.
I perceive this as another wireless transmission mechanism that can be targeted to a specific user, like lasers. Such transmissions reduce the ability to intercept signals without the receiver knowing that the signal has been intercepted. Certainly, part of these streams should include a timestamp with any strength/time variances should set off alarms.
If there are other cryptographic properties to this TRA application, it isn't clear from the article. Bouncing the sound back to the source doesn't seem to be useful in a data transmission application.
For any stereo-/audio-philes reading this, remember how good a Linn Sondek turntable makes your vinyl records sound. The turntable captures a larger part of the analog data off the vinyl and works WITH the harmonics, rather than trying to squelch them. From http://www.patricktaylor.com/linn-sondek-lp12
"The grooves of a vinyl record contain more information than a compact disc of the same recording. A CD may hold more minutes of music, but what is contained on the disc is a digital representation of the musical signal. Missing are nuances that give music depth as well as the sense of a live musical experience (Popular Mechanics Magazine, 2003)."
I wonder if the unordered material might be suspended in an electro-viscous fluid. The sender could removed the material and change the viscosity of the fluid to something like water. It could be shaken like a snow globe and then 'frozen' in place through a reapplication of current. This would be kind of like changing keys with a brownian motion.
Sounds like this could be used in either of two ways.
In one use the sound is focused to a specific point - imagine sending private messages to one person in a room unintelligable to others in the same room. Intercepting sound at a different point in the room lacks enough information to listen to the message. If you consider the listener's location as the "key" then this looks something like encryption.
In another use the disordered medium is the key used to decypher the message - equivalent to an acoustic hologram. The usual problems with any shared-secret system apply. As to whether this has any sort of practical advantage - not at all clear.
Whether or not it is "cryptography" per se is almost irrelevent. Anything that cost-effectively helps prevent a secret from being spilled may be important, depending on the importance of the secret. Steganography does have its place. Think Submarines and Spys. The described system may work as a way of concealing that a message was even sent.
Time-reversal stuff is pretty standard; what surprised most scientists is how easy it is to do. An interesting experiment I remember from a few years ago had a human speak on one side of a randomly placed "forest" of plastic bars and cones, on the other side of which was an array of microphones. When they played back the sound through speakers in the same places as the microphones, the original voice was clearly audible in the original speaker's position, even though the sound was gobbledygook on the microphone array side.
They found that it didn't really take that many microphones to get an intelligible playback -- but that probably says a lot more about the low entropy of the human voice than anything about the scrambling/descrambling.
It is possible with that sort of experimental setup to map out a response (at the microphone array) for any particular waveform -- so you could then generate intelligible speech at the original source point by adding all the various transformed waveforms together in the microphone/speaker array.
The main problem with this sort of technique for transmimtting data is that it requires the receiver to send out a signal first, so you can map the intervening medium -- it can't be a true broadcast.
If Bob is on a submarine out in the Pacific somewhere, and Alice has a large set of microphones up and down the west coast of North America, then Alice can send Bob a nearly uneavesdroppable signal -- if, and only if, Bob sends Alice a signal first. Bob has to broadcast a signal which can be used by Alice (using her large-scale array of microphones) to construct a map of the intervening medium -- or more accurately, to construct a mapping function that allows the Alice (using the transmitters in the same places as the microphones) to then send data back to the Bob, with a waveform which will basically converge on the receiver with little or no spatial spillover.
The only ways that would be practically useful are
- if Bob can transmit the locator broadcast, receive a reply, and get away before anyone can catch him
- Bob can transmit the locator broadcast at some earlier time, and then come back later; this assumes (a) that the mapping function doesn't change drastically with time and (b) Bob can get back to the same position with enough accuracy that he's inside the focussed zone
And even then, it's almost certainly possible to eavesdrop on the signal -- as long as you catch both sides of the exchange (the original establishing signal plus the resonponse).
This sort of technique is probably much more useful as a way of using low power transmissions in an electromagnetically noisy (and echo-y) area. I believe this is an area of active research in wireless communications.
This is potentially useful as steganography; cryptographically, I'm not sure.
great for driving people nuts too. Who else hears the voices, or is it just me?
"I simply don't have the science to evaluate this claim"
Life experience be more relevant I would have thought. It's called "whispering" ain't it ?
Can't really think of any other situation where you would be using sound to communicate messages with security implications. Distance and practicality don't really support it's use over electronic/radio/light for anything else.
A "time reversal mirror" is simply a transducer array. "Hyperfocused" simply means "focused" as with an optical lens or a sonic array. An array can be used with a lens to further focus sound waves.
Once you translate their hyperbolic speech, it simply reduces to producing sound beams using arrays and lenses. This is how all focused waves are produced.
For those who think that sound beams will provide privacy, I refer you to the issues with infrared, microwave and many other sorts of electromagnetic beams and surveillance. Or you can just think of a megaphone and how private that is.
Seems like you could "brute force" this if you captured all original input signals and then put a computer to trying a large number of phase relationships between them until the result made sense.
This attempt to apply a principle of physics to cryptography and security has the typical flaw of most such attempts:
It assumes that no counter-attack can be developed also using physics. The poster above points out such a possible counter attack using 'brute force'.
The same flaw occurs in so-called quantum cryptography. They claim it is secure because you cannot intercept a bit without changing it, but we really have no way of knowing that there can never be found a way to do so. Especially in the field of quantum mechanics, much is not known.
With OTP, you can prove mathmatically that it is secure. But with any method of 'secure' transmission of data based on a physics, you can never prove that it is secure, because we do not know everything about physics.
It seems like the ability to put microphones all around, capture garbled sound from a whole room, and then play the sounds back out and hear what was being said at each point is enormously valuable for surveillance. Wouldn't it be cool to combine this technique with a surveilance camera and then go back and listen to all the interesting private conversations at a coctail party or whatever?
This is everyday science for your average audio professional who designs and stack large-scale pa-systems. What you are trying to achive is even dispersion for the crowd, and so little sound-spill as possible where you don't want the sound to hit. If the sound bounches of a wall an hits the crowd with a greater time difference between the reflected sound, and the direct sound from the speakers of 50ms, you get an audible echo, something considered really bad. By using speakers with a narrow dispersion pattern, you can focus the sound and steer it away from locations where you don't want it to hit. If you could make a speaker with a verry tight dispersion pattern, something like 1x1 degrees, you could theoretically "hit" a single person over a large distance, but the aiming of the system would be verry difficult if the reciver vas moving, and if you missed the intended reciver, and hit someone else, they would hear what you were transmitting.
This concept is covered in the current of Make Magazine. I don't have a link, it might not be available on the web yet, but I've seen it in the print version.
>great for driving people nuts too. Who else hears the voices, or is it just me?
Oooo, or let's say you're going to be in, oh, a debate of some sort and don't want to have a stereo boom box taped to your back so your handlers can feed you something like competent answers....
"With OTP, you can prove mathmatically that it is secure. But with any method of 'secure' transmission of data based on a physics, you can never prove that it is secure, because we do not know everything about physics."
Yes, and most cryptosystems are going to fail if some novel new crack is found (Note that I don't see this particular thing as any more than a hypothetical way to have a single line of wireless communication). Frankly, the problem with OTP is that you somehow need to safely distribute the pads. In many cases, this is as difficult a problem as sending the message without any interception. In the rare cases where you have 2 individuals that infrequently physically meet, but need secure communications between those points, it has some use. When you expand the number of individuals that want to intercommunicate, the problem swiftly becomes intractable. If there are problems with physically meeting, the problem swiftly becomes intractable. If meetings are frequent, then you might as well just communicate the messsages at the meeting time. So the main problem with OTP is that it dosen't solve most cryptological problems.
My company, C-Lab, looked into this for a customer building airport paging systems a few years back, and got it working fairly well. The trouble is knowing where your target is, mainly, and also getting all the speakers installed into very well known locations -- installers tend not to be skilled enough for this. You need fraction of a wavelength accuracy for the highest frequency involved to make it work nicely. Then you can use lots of dispersed, low power, say 1 watt (electric, acoustic output is less) sources to lift someone right out of their seat from quite a distance. It can be most impressive in a boardroom demonstration setting! Everyone else can barely hear it, if at all, or what they can hear is garbled due to the multipath delays. We ditched the idea as impractical, even though possible, due to the above limitations. Too bad, it sure was fun to play with! Of course, if you can also characterize all the room reflections (and time-reverse them) you can do this even better, but people moving around really affect this in most real life situations, as they absorb sound and affect acoustics quite a lot.
So Alice has to record Bob's signature sound and play it back (backwards) to send Bob a message, assuming he's still in the same place. Noone else will hear anything above the background noise. That is a fascinating concept.
The problem is that Alice's backwards playback session will send signals to anywhere else in the vicinity that had some noise happen concurrently with Bob's broadcast, because those sounds will be recorded on Alice's recordings.
If Eve could arrange to emit a sound at the same time as Bob emits his, then she'll have copies of Alice's messages sent directly to her. It doesn't seem like Alice or Bob could detect this.
Alternatively, if Alice sends messages frequently enough, Eve could roam around with her ears open looking for a spot where some sound occurred at the same time Bob sent his signature.
Not an adequate substitute for solid mathematical encryption. An added layer of protection? Maybe.
Schneier.com is a personal website. Opinions expressed are not necessarily those of BT.