by DAVID KAHN
Even in financial dealings, banks seem to prefer to rely upon codes and ordinary precautions. The system works well enough, if the experience of 26-year-old David Hermoni is any guide. Hermoni, one of three employees of the Hollandsche Bank Unie in Haifa who knew the bank’s private code, opened two accounts in a Zurich bank on September 1, 1958, on his way back from a vacation. The accounts bore only numbers, not his name. He then cabled three New York banks (the Irving Trust Company, the Manufacturers Trust Company, and First National City Bank) in the private code, instructing them to transfer $229,988 to his two accounts. After calling in sick at his job, he flew to Zurich, identified himself to the Swiss Bank Corporation as the owner of an account, and withdrew $150,000. He then went to Credit Suisse and drew out $50,000 from his other account. But when he returned to the Swiss Bank to get another $25,000, he was arrested; a confirmation cable to Haifa from one of the New York banks had tripped him up. Since things like this could happen as well with cipher systems, banks find it unnecessary to invest in them. The International Monetary Fund, on the other hand, employed Mrs. Friedman soon after it was created to set up an elaborate cryptographic system, based upon the one-time tape; the fund built a big safe to hold all its tapes. But its situation differs from those where banks deal with private transactions; the fund’s activities may have international repercussions, and interested governments may seek to discover its plans so that they may take self-advantageous economic action.
Commercial secrecy has had a moderate success in one field: telephonic communication. The convenience and universal use of the telephone and the prevalent fear—if not the actual prevalence—of wiretapping or switchboard eavesdropping has led some businessmen to buy scramblers. Excluding the installations constructed by the large communications companies to protect their radiotelephone traffic, at least three companies in the United States make scramblers. Delcon Corporation of Palo Alto, California, produces several kinds, from a simple, portable telephonelike device that fits, hand-held, over an ordinary telephone to scramble the outgoing words and descramble the incoming, to more elaborate radio-scrambler attachments. They are inverters and the company describes the effect as an “unintelligible jargon which can be vaguely identified as the sound of an incomprehensible foreign language, similar to the effect of a phonograph record player in reverse.” Delcon provides different ciphony arrangements for different customers—presumably different inversion points. Among those customers, oil companies and mining firms again lead the pack. Prospecting teams will carry a scrambler with them so that they can report back from the field without fear that a wiretap will reveal their information. The Shell Oil Company used them, for example, for talks with drillers and lease buyers. The helicopters that scout for fish report the locations of large schools via Delcon’s radio scramblers so that rivals will not learn where the good fishing is. The ships themselves scramble their price discussions with canneries so competitors will not underbid them. During their multimillion-dollar proxy battle with Allan P. Kirby for control of the $6 billion Alleghany Corporation in 1961, brothers John and Clint Murchison of Texas telephoned one another using portable scramblers. Police use the radio scramblers on stake-outs so that criminals with police radios will not know that they are being watched. The prices range from about $150 for the portable hand-scrambler to $450 for the radio scrambler.
The Westrex Company, a New York City division of Litton Systems, Incorporated, makes two inverters which differ only in some minor technical details. Both accept a speech band of from 250 to 2,750 cycles per second and invert it about the midpoint of 1,500 cycles. El Al Israel Airlines used the Westrex system to provide privacy for radiotelephone calls from its airplanes high over the Atlantic. Most elaborate—and most expensive—of the commercial scramblers are those of Lynch Communication Systems, Incorporated, of San Francisco. Its E-7 is a 319-pound band-splitter offering 233 combinations; several of these have reportedly been sold to some Latin American countries and to some communications firms. Lynch also makes a 71-pound inverter, the B-69, which appears to preserve the quality and intelligibility of the speech much better than the other inverters.
The most primitive form of human graphic communication—pictures—resisted subjugation to the methods of secrecy much longer than its younger brother, writing. Images had to await the invention of technical means of reproduction before they could be distorted or scrambled for secrecy. Cryptographic literature records a few rare cases where spies clandestinely transmitted plans of fortifications by disguising them as parts of a drawing of a butterfly or a landscape. But this belongs to steganography.
Cryptoeidography (from the Greek “eidos,” “form”) encompasses two basic ways of making pictures secret. One is based upon optics; it takes the image directly and distorts it. The other way is based upon electricity; it distorts, not the image, but an electrical current that represents it. Whether the former is akin to code or not, the latter resembles cipher and, in respect of its clear analogy to ciphony, is called “cifax” (from “cipher”+ta shortening of “facsimile”).
It might seem at first that the optical systems would be the older. But though lenses for distorting have been available since at least the time of Anton van Leeuwenhoek, there was for a long time no way of recording the distorted image. And when Louis Daguerre devised such a method, it soon became clear that no amount of viewing an out-of-focus photograph through correcting lenses would reproduce the original external object with clarity. Since this would mean that an encoded picture could never be decoded, no systems based on this method were devised. (Microphotography served in communications, but that is a method of steganography.)
Perhaps the first cryptoeidographic system to employ classical optical principles was a stereoscopic one. A stereoscopic photograph actually consists of two pictures taken simultaneously by two cameras a small distance apart; when the two images are viewed in a special holder together, the eyes combine them to give a three-dimensional appearance to what they see. The two films differ only in very slight displacements of the images of the objects shown—displacements ranging from only a fraction of a millimeter to about three millimeters. “Because the significant stereo differences are those of lateral displacement, it is easy to imitate them in hand work,” wrote Herbert C. McKay, director of the Stereo Guild. A plaintext is scattered through a cover-text—which may be intelligible, but need not be—consisting of many nonsignificants. The encipherer typewrites this cover-text on a sheet of paper. He types it a second time on another sheet, omitting the letters of the plaintext, then shifts the paper in his typewriter ever so slightly and fills in the significant letters of the plaintext. He sends the two sheets by different routes to the recipient, who inserts them in his stereoscope. The plaintext optically leaps up off the page at him in relief. “Because similar irregularities will not affect stereo relief,” explained McKay, “it is possible to introduce random irregularities which will not change the stereo appearance but will positively prevent any attempt to read the message by measurement of spacing.” Though the system labors under some practical disadvantages—both messages must reach the recipient; the interceptor needs only a stereoscope to solve them—it has an intrinsic theoretical interest. A later system based on classical optical principles employs many tiny lenses to separate the plaintext image into small portions and then to rotate these portions out of alignment with one another. This system was first devised in 1960 at Bausch & Lomb, perhaps because it could not work efficiently until plastics technology developed a method of molding in a single piece the scores of lenticles needed.
A more sophisticated form of optical cryptoeidography emerged with the evolution of what came to be called “fiber optics.” It had long been known that light would travel in a curved path by repeatedly reflecting itself from the inner surface of a thin, curved conductor, which could be water or glass or other substances. But until the 1950s it had never been possible to transmit a picture along such a path. Then Dr. Narinder Singh Kapan
y, then in his late 20s and at the University of Rochester, bunched many glass fibers, each about a thousandth of an inch thick, into a bundle. Each of these hair-thin “light pipes” picked up a point of light from an illuminated image and transmitted it faithfully to the opposite end of the bundle. Here the fibers, which occupied the same relative position at both ends, reproduced the image in the form of hundreds of thousands of microscopic points of light and dark. Kapany realized that if the two ends of the bundle were not alike, if the fibers occupying, say, the edges of the input face occupied the center of the output face, then the emerging image would be scrambled. To decode, the image need only be sent backward through the same or an identical bundle. He tested the idea, and it worked. A picture of numbers and lines emerged from a bundle of about a quarter of a million fibers as an absolutely random and shapeless grouping of black and white dots. Its decode appeared grainy—as do all such images—with some “holes” in parts of the numbers and lines, but fully intelligible.
In this form, however, the fiber-optic coder labored under several practical difficulties. To make one, the fibers of a bundle are scrambled in the middle of the bundle, which is cut at that point. The two halves then serve as the encoder-decoder pair. It is, however, extremely difficult to reproduce a particular scramble; moreover, the loss of fiber material during the cut decreases the accuracy of the decode. To eliminate these problems, Robert J. Meltzer of Bausch & Lomb, Incorporated, dissected the plaintext image not with individual fibers but with many small bundles of fibers about a tenth of an inch thick. They pick up the light of the image at the input face, offset it through a scramble of the bundles, and emit it in jumbled form at the output face.
For once there appeared to be a fair potential in business for a coding device. If someone finds a bank passbook or an identification card with the owner’s signature on it, the finder can forge the signature to withdraw money or gain improper access somewhere. Personal credit companies reportedly have sustained considerable financial losses in this way. But if the signature were encoded, the finder would be virtually unable to reconstruct it. Three companies have offered signature-encoding systems to industry: the Le-Febure Corporation of Cedar Rapids, Iowa, a subsidiary of Craig Systems, Incorporated, with an Autho-Visor system, R.C.A., with its Signa-Guard systems, and Bausch & Lomb. In general, they convert the signatures into broken wavy lines, rather like a highly magnified fingerprint. But despite the apparently bright prospects, sales of the systems—which cost several thousand dollars—have not mounted very high.
Cifax systems have never succeeded commercially, either. Probably the most primitive are those for enciphering the telautograph, which reproduces handwriting at a distance. Gilbert S. Vernam used a gear mechanism on the telautograph to add a circular rotation to the original motion of the hand-guided pen, producing a scrawl. The decipherer followed the scrawl, and his mechanism, subtracting the rotation, traced out the original writing. Vernam also invented a mechanism that analyzed a picture into shades of white, light gray, dark gray, and black, converted them into holes on punched tape, and enciphered them with a keytape according to the Vernam principle.
Cifax took a major step forward with the invention of wirephoto and radiophoto. A photograph is mounted on a rotating drum. A photoelectric cell scans the entire surface of the picture and converts the gradations of gray into a fluctuating electric current for long-distance transmission. The brighter the spot of the photograph being scrutinized by the electric eye, the more current it will send out. At the receiving instrument, light-sensitive paper on a similar drum turns beneath a light source that shines brighter or dimmer in ratio to the electric current; gradually the entire sheet is exposed. Normally, the two drums rotate at the same constant speed; if they are out of synchronization, distorted images result. The French engineer who invented this system, Edouard Belin, seized upon this weakness and made it a cryptoeido-graphic strength. He rotated the drums at irregular intervals according to a prearranged key. As long as the sender and receiver ran according to the same pattern, the deciphered images appeared normal. But an interceptor would get only a blur of smudges, streaks and white spots.
The author’s signature scrambled by fiber optics
Belin’s was the best-known system for enciphering still pictures, but many others were invented, for the electric current of wirephoto could be deformed in just as many ways as the electric current of a telephone. Engineers could invert it. They could divide it into frequency bands—representing here levels of brightness instead of levels of voice—and substitute one for another. They could subject it to a time-division scramble, or bury it in noise and extract it at the receiver. But few of these methods were ever used: no one needed them.
Then came television, and in the 1950s in the United States a great battle over subscription television, also called “pay-TV,” “toll-TV,” or simply “fee-vee.” In pay-TV viewers pay a nominal sum to see first-run movies, Broadway plays, sports events, opera, ballet, and other attractions not normally on television—and to see them uninterrupted by commercials. Apart from the struggle within the industry as to whether the Federal Communications Commission should license any kind of pay-TV at all, the fee-vee partisans disputed among themselves as to the best kind of subscription. Some favored broadcasting the subscription programs scrambled so that only subscribers with a decoder attached to their television sets could get a comprehensible picture. The subscribers would pay an installation and monthly rental fee for the decoder and an additional fee for each program they watch. Other subscription TV firms proposed bringing the programs into the subscribing homes by wire. In urging wired TV before the F.C.C., the Jerrold Electronic Corporation of Philadelphia showed how easy it would be for a pay-TV bootlegger to solve any broadcast scramble and to sell information or equipment to unscramble it to the public. Jerrold’s two reports, by Donald Kirk, Jr., its vice president and director of research, comprised perhaps the first discussion in cryptology of TV cryptanalysis.
Television cifax operates upon certain characteristics of the television signal, and it is no more possible to understand it without knowing how television works than it is to understand codes and ciphers without knowing what letters and words are. The TV camera converts the light and dark parts of the image focused by its lens upon a photosensitive surface into proportional fluctuations of electric current, which is transmitted as a radio wave. The brighter the spot, the greater the amplitude of this wave. The television receiver transforms the fluctuations of this wave into equivalent fluctuations of a beam of electrons directed at the phosphorescent face of the picture tube. The greater the amplitude of the incoming wave, the heavier the beam of electrons pumped out by the receiver—and consequently the brighter that spot on the picture tube. The camera scans the photosensitive surface 30 times a second, the close succession of pictures giving the impression of motion. Naturally, the electron beam in the receiver must sweep in exact synchronization with the camera, and to assure this the transmitter sends out at the proper times a pulse that tells the receiver, “Now start again at the left-hand edge of the picture and begin sweeping toward the right at the predetermined rate.” American television divides the picture into 525 horizontal lines. The horizontal position pulse is transmitted for each of these lines, or about 1,500 times a second. To further assure the synchronization, the transmitter also sends out a vertical position pulse that tells the receiver, “Now start again at the top of the picture.”
This system makes several variables available to cifax. The most obvious is the basic video signal—the one determining the amount of light, or brightness, of the parts of the picture. This signal is analogous to the one that carries the frequency of a voice, and cifax may deform it just as scramblers distort the frequency signal. Simplest of all is inversion. The blacks become whites and vice versa, and gray tones invert around a midpoint. A video band-splitting would divide TV brightness into five groups, ranging from dark to light, and replace, say, group 1 with group
4, group 2 with group 3, and so on. If these assignments remained unchanged, Kirk observed, the result would constitute a kind of cifax monalphabetic substitution. But he noted that a one-time system, affording perfect security, was theoretically possible. As the transmitter scans across a line of the image, it discerns about 300 individual spots of light. Filters would discriminate each of these into one of television’s approximately 25 levels of brightness. Then a one-time key would control the substitution of one level for another. Between the extremes of monalphabeticity and one-time keys lay many possibilities, Kirk wrote. “In a conventional LP [long-playing] record of the 33 1⁄3 r.p.m. type if one considers that the frequency response is out to 5,000 cycles and that this record plays for some 30 minutes, then one might conceivably get 10,000,000 pieces of coding information on the record. Now if this record were to last for one month and during that month there were to be 250 hours of television programming (this is about one million seconds of television programming), then one can see that this information stored on an LP record might be used to shift the coding signal some ten times each second…. Of course, it is unlikely that one could record as much information as has been listed here on an LP record and expect it to stay synchronized [with the LP record used at the transmitter to encipher the signal] for 250 hours of television programming.”
