Optical Cryptography

chill writes: "In Cryptonomicon, Neil Stephenson wrote about Bell Labs' research into using static, or chaotic signals to mask communications. A message would be generated, then the signal masked in noise. Someone on the other end would subtract out the noise to get the signal. Works great if both ends have the exact same noise. Now, Jia-ming Liu, professor of electrical engineering at UCLA, is giving a presentation on doing essentially the same thing using OC-48 (2.5 Gbps) optical circuits. The presentation will be at the upcoming Optical Fiber Communications Conference and Exhibit. There is an article covering this and some other nice advances in optical over in Wired."

Comments

Seems like a waste of noise...

[b0r0din]

Maybe I'm completely off here, but if you're using noise interference, wouldn't that be sort of wasting bandwidth? This is a cool technology, I wonder if there would be a way to mask a signal and at the same time run multiple signals, so you could essentially split the information through a long pipe (like the laser) using the chaotic noise, and each would be able to be filtered out (at some sort of router) and sent to various places accordingly. Seems it would be much more efficient to carry information that way.

Use BWT instead of LZ for even more diffusion

[yerricde]

How does one hide messages in reandom noise, though? Would it work to LZ-compress them, to make them appear random?

LZ+Huffman (i.e. deflate, the core of gzip and pkzip) works, but you get more compression in a Burrows-Wheeler based scheme such as bzip2. More compression => more entropy per coded symbol => more resistance to known plaintext attacks.

Nulls.

[TheSHAD0W]

This technique is actually very old, though it wasn't used bit by bit. You're inserting null terms into the cypher stream. Prior to modern cryptological methods nulls were fairly popular, but the technique has fallen into disuse because of its increasing the message size, and because 1:1 stream cyphers are SO much more convenient. Besides, the new cryptosystems are unbreakable, right? Right?

Even having a small multiple of nulls to significant elements increases the complexity of calculation exponentially. For example, a 1:1 proportion of null bits in 512-bit blocks. The result is a 1024-bit blocked key stream. You can't do any sort of intelligent analysis of the stream unless you can figure out which bits are significant, and there are 2^512 possible permutations of significant and garbage bits for each block.

Chaotic crypto crackable, OTPs not

[billstewart]

Most of the chaotic cryptosystems people have tried to design have been crackable, and cracked. Perhaps there's something about this one that's different, but just because something's relatively hard to predict doesn't make it impossible for people who are far better at math than I am.

By contrast, a theoretical one-time pad is theoretically provably uncrackable - if you really do have uncorrelated random bits for your pad, and you really only use them once, it's perfectly secure, and even knowing N-1 bits of a message tells you nothing about the other bit. In practice, source of random numbers aren't always perfect, and sometimes people cheat and reuse pads - the NSA's "Venona" crack of Soviet crypto primarily succeeded due to rampant reuse of pads by sloppy crypto users, though I think they also found some non-randomness in the pads that they could exploit a bit. But this optical system guarantees that if you know the initial conditions, you can use the first N-1 bits of a message to predict the next one, and sometimes you may be able to deduce those initial conditions closely enough to crack the system.