Although security has generally been a success story for optics, staying technically ahead of criminals is an increasingly difficult task. Counterfeit holograms, for instance, are becoming commonplace. However, researchers from the University of Connecticut (UConn; Storrs, CT) and Thomson CSF (Orsay, France) believe they have come up with an encoding method that should be almost completely undecipherable. The system has random phase codes in both the input and Fourier planes. Even if the codes are not very complex, cracking both of them simultaneously to retrieve the signal could potentially take centuries.
This work is an extension of a technique that Bahram Javidi, a professor at UConn, developed with Joseph Horner of Hanscom Air Force Base (MA). Javidi and Horner found that, by using a transparent phase mask placed over a photograph or signature, they could verify that an identification or credit card was genuine and had not been tampered with. When the card was illuminated with coherent light, the phase signal from it could be compared to a reference phase signal within an optical processor. If the two signals interfered to produce a correlation peak then the card would be verified.
Though it would not be easy, the coded phase mask used in that approach could still be forged. Javidi, working with Philippe Refregier of Thomson CSF, discovered that adding a second phase mask could make the technology insurmountable. The encoding system consists of two lenses and two phase masks. The data to be encoded pass through the first lens, encounter the first mask, are inverse-transformed by the second lens, then pass through the second mask. The result is an image that looks essentially like white noise, both in phase and amplitude.
To decode the information, it must be passed through the inverse of the optical system used to code it, which means that the counterfeiter would have to know the exact composition of both phase masks. Even for a relatively simple case, "guessing" could take a prohibitively long time. A single, binary 10 × 10 phase mask could be in any one of 1030 states. Together, two similar phase masks have 1060 possible states. Even if it were possible to do 1012 correlations per second, it would take on the order of 1040 years to go through all the possible permutations.
Even with spectacular improvements in computing speed, similar (but not identical) masks giving partial results, more restrictive encryption codes, and blind luck, this technique should be secure. And, if it appears that a system has been compromised, the user could simply change the phase masks, perhaps moving to a larger number of pixels or a multilevel system to dramatically increase the difficulty of the problem.
As with many optical-processing techniques, alignment could be problematic. Fourier transform-based systems are generally shift invariant, but rotation errors are potentially critical. However, once aligned, two phase-only spatial light modulators could do the job, allowing the phase mask configuration in the encoding and decoding to be changed regularly or even to be different for each card.
Of course, the system could still be beaten if counterfeiters simply stole the mask configuration codes or got hold of decoding equipment. But this would be true for almost any system. If the double-encoder system could make the cards themselves secure, researchers would have plenty of time to develop optical solutions to other security issues.