by DAVID KAHN
Engstrom was succeeded by Louis W. Tordella, who, at 47, was the youngest to be named deputy director and who has held the post the longest. He taught mathematics at Loyola University in Chicago and at Illinois University from 1935 to 1942, where he obtained a Ph.D. in mathematics. He served in the Navy, presumably as a cryptanalyst, during the war; afterwards, he remained with the Defense Department. His mathematical specialties are algebra, group theory, and classical number theory. His appointment may represent a trend away from high-powered managers brought in from the outside and toward a policy of permanent career administrators to assure stability and continuity despite the political changes of the heads of the agency.
The agency that they run is divided into three operating divisions and a group of supporting administrative units. The three operating divisions are the Office of Research and Development, with about 2,000 employees, the Office of Communications Security, with about 1,500, and the Office of Production, with more than 7,500. The main supporting units are the Office of Personnel Services, which recruits and hires, the Office of Training Services, fourth largest unit in the agency, and the Office of Security Services, which maintains physical and personnel security, reviewing the background investigations of prospective employees, giving lie detector tests, and granting—or refusing, or revoking—security clearances. Smaller supporting units include the offices of the director, the comptroller, and the adjutant general, the inspector general, counsel, and the library, headed by Dr. John Sanford, which has a superb collection of works on cryptology and of up-to-date reference texts (needed by the cryptanalysts for probable words) and more than 600 mathematical publications in English, Chinese, French, German, Portuguese, Russian, and Spanish. At least once its copy of Shakespeare served a non-literary purpose when a cryptanalyst, cracking a spy cipher, recognized the first words of the key as a quotation from the Bard, rushed to the library, dug it out, and effortlessly reduced the cryptogram to plaintext.
Without people, of course, N.S.A. could not even exist, and getting and keeping personnel appears to be one of the chief continuing problems facing the agency. This is the task of the Office of Personnel Services.
The first part of the problem is getting them in. N.S.A. depends heavily on scientists and engineers. Their skills are in short supply, and competition for them is intense. Consequently, N.S.A. actively conducts a nationwide recruitment program, directed primarily at young college graduates because government salaries at their levels do not differ from industry’s so markedly as with experienced men. Scientists in colleges throughout the country refer promising graduates to N.S.A. and recommend N.S.A. to them; their high professional caliber has been a major factor in the recruitment program’s success. N.S.A. recruiters tell the prospective employees about opportunities to climb either a supervisory or a technical career ladder. As Canine put it, “There is a marshal’s baton in each employee’s knapsack when he starts out”—a metaphor that may be somewhat felicitous for cryptology in view of the story (though it is almost certainly apocryphal) that the skytale, which the Spartan commander always wore at his waist, became a symbol of authority that is today the swagger stick of the field officer and the baton of the marshal.
Applicants must pass the Professional Qualification Test and then the eagle-eyed scrutiny of a security check. So rigorous is this that, despite the agency’s urgent need for scientific talent, five out of every six applicants are rejected. From the nation’s college graduating classes of 1956, N.S.A. hired between 250 and 300 young men and women. Not all were scientific or engineering or even language personnel. N.S.A. also needs liberal arts graduates to work in its library and other support facilities. And the graduate’s experience need not be directly related to his work, since the agency will train him for what it needs. Often it prefers to do this—taking a French major, for example, and teaching her Russian. Cryptanalysis, must, of course, be taught, and one member of the Office of Training Services, Lambros D. Callimahos, is revising the Friedman Military Cryptanalysis texts to bring them up to date.
As a consequence of this intensive recruitment, many N.S.A.ers are young. The agency regards all appointments as permanent, with no probationary or temporary positions, and it does its best to hold its hard-won new employees. It arranges excursions to New York. It stages hobby shows. In the evenings the sedate classrooms of the N.S.A. School, which during the day echo to the fricative paradigms of Russian verbs and dry-as-dust exercises in symmetry of position, resound to the hot rhythms of the rhumba and, perhaps, the twist, as N.S.A.ers take dance courses. Earnest young scientists, who cannot publish their highly classified work in the usual scientific publications, satisfy their yearnings for professional recognition by writing for the N.S.A. Technical Journal.
Yet personnel turnover is high. Young men and women who are just finding themselves change jobs. Industry offers more money. And always there is the oppressive atmosphere of security. Despite the picnics and the dances, the compartmentalization of offices and the restrictions of movement tend to limit the romantic aspirations of the girls in their early twenties. One girl blamed the omnipresent secrecy for keeping her from learning that a man with whom she had a love affair was married; while this was undoubtedly more her fault than the agency’s, it does indicate the resentment that the security breeds. And the threat of sudden, instant dismissal, without recourse to a hearing or review and without any confrontation of any person who might have accused the employee, justly or unjustly, of something that the director feels might jeopardize N.S.A. security, cannot enhance morale.
The factors that outweigh all these, however, and that largely enable N.S.A. to retain its staff, are patriotism and the opportunity to serve. These provide spiritual satisfactions that money cannot buy.
The agency runs itself in accordance with modern principles of management. Its interne program seeks to develop civilian employees for high management posts. It promotes from within, and moves personnel from post to post to broaden them. It maintains a suggestion program that pays out hundreds of dollars for good ideas. It offers instruction in dictographs for supervisors, appraises its paperwork, and works hard at keeping itself as efficient and streamlined as possible.
The cutting edge of cryptologic progress in the United States is N.S.A.’s Office of Research and Development, or R/D. Solomon Kullback, one of the three cryptanalysts that Friedman hired in 1931, served as its head in the early 1950s. In 1957, Howard H. Campaigne, a Ph.D. in mathematics specializing in statistics and hypergroups, became head of the mathematical section at the age of 47. His assistant is Dr. Walter W. Jacobs, a mathematical statistician who had previously served in the Office of Production.
R/D is divided into three sections called REMP, STED, and RADE. REMP—the term stands for “Research, Engineering, Mathematics, Physics”—conducts basic cryptanalytical research. It ransacks the domains of statistics and higher algebra for ever more sensitive and more powerful tests to solve complex ciphers. It attacks difficult foreign cryptosystems to devise new techniques of solution; any intelligence obtained is, so far as R/D is concerned, a by-product of this search. It advises other N.S.A. divisions on problems involving new methods. It works intensively to improve computer applications to cryptology. Engineers and physicists seek increases in computer speed and data-handling capacity by transistor circuitry, short-pulse techniques, time-sharing, and magnetic memories. A recent effort involved eliminating speed-inhibiting factors from such memories. REMP uses computers to design computers, and engineers working on peripheral components, such as line printers and punched-card inputs, must strain to keep up with basic technology. N.S.A. leads even such firms as I.B.M. and Remington Rand in important areas of computer development, such as time-sharing, and industry has adopted many N.S.A.-designed features.
The second section, STED (for “Standard Technical Equipment Development”) conducts basic cryptographic research. It looks for new principles of encipherment. It ascertains whether new developments in technology,
such as the transistor and the tunnel diode, have cryptographic applications. Using such esoteric tools as Galois field theory, stochastic processes, and group, matrix, and number theory, it will construct a mathematical model of a proposed cipher machine and will simulate its operation on a computer, thus producing the cipher without having to build the hardware. Rotor principles have often been tested for cryptographic strength in this way. It devises audio scramblers, from the ultra-secure types for high officials to the walkie-talkies of platoon commanders, as well as video scramblers for reconnaissance television and for facsimile. Their development involves sciences from metallurgy to optics, as well as techniques—important in miniaturization—from printed circuits to ferro-resonance.
R/D’s third section, RADE (for “Research And DEvelopment”), conducts basic transmission research, going deeply into such matters as the interaction of electromagnetic radiation and matter. It aims both at increasing the sensitivity of American intercepting receivers and the security of American transmission methods. N.S.A. radios operate at the extreme limits of radio frequencies and involve all types of electromagnetic emanations. Its listening posts require both panoramic receivers to scan the entire frequency spectrum and single-frequency receivers with a high degree of stability that will not drift off a signal. RADE strives constantly for antenna arrays that will accentuate the signal and eliminate atmospheric interference and circuit noise so as to pick up even the weakest radio messages. It improves direction-finding apparatus and devises radio fingerprinting apparatus. And it looks into new techniques of communication, such as methods that spread a transmission over so broad a frequency spectrum that anyone listening on one frequency band would hear only a faint crackle like static. These may themselves afford some security—at least until the enemy’s technology catches up. Presumably it is investigating the possibility of sending messages by laser.
In addition, N.S.A. engages in some basic communications research in the broadest possible sense. The flow of impulses through a computer’s circuits constitutes a study in communication, and N.S.A. mathematicians investigate it. They use the tools of the new field of information theory to look into other problems—compression of maximum information into a minimum bandwidth, expected percentages of errors, rates of transmission, pattern recognition. N.S.A. physicists study modern quantum theory of many-body systems, superconductivity, magnetic resonance, the electromagnetic properties of solids, and the scattering effect of the ionized region of the troposphere for possible application to communications. Language itself is dissected phonetically, phonemically, grammatically, logically, semantically, historically, statistically, and comparatively. These studies result in one of N.S.A.’s few unrestricted products: dictionaries and grammars of more recondite tongues, such as the 429-page Vietnamese-English Vocabulary issued by the Office of Training Services, the Romanian-English Dictionary, prepared by N.S.A.’s 762 Dictionary Unit, and A Grammar of the Bulgarian Language. R/D’s research differs from that carried out by the Institute for Defense Analyses’ Communication Research Division at Princeton in generally being rather more applied in nature. I.D.A. research is freer, more “far out.”
Smallest of N.S.A.’s three operating divisions—and the only one whose duties are publicly acknowledged—is the Office of Communications Security, or COMSEC. It is responsible for the protection of secret American government communications. Consequently it prescribes or approves the systems each department must use and how they must use them. It furnishes some machines itself and lets contracts for the others. It promulgates the national cryptosecurity doctrine and supervises its execution.
“All cryptographic material (including cryptographic equipment, instructions, spare parts, and associated materials for the Armed Forces) is produced by, or procured under, the direction of N.S.A.,” states an Air Force manual. The same must be true for the Army, the Navy, and the State Department. COMSEC standardizes as much of American cryptography as practicable, down to the short titles of communications security publications. Thus the Air Force Communications Security Manual 2, formerly known as AFCOMSECM-2, is now listed as AFKAG-2. COMSEC prepares courses of instruction for new cryptographic equipment and issues regulations for its operation, presumably mandating such matters as the when and how of primary and secondary key changes. For interdepartmental and presidential communications, it probably produces the keys—rotor wirings, lists of positions, one-time tapes. Keys for communications wholly within, say, the Air Force are presumably produced by its own cryptographic agencies.
COMSEC, drawing upon R/D’s STED, devises new systems of encipherment and embodies them in new mechanisms. It works closely with potential users, such as the State Department, to make sure that the equipment fits the user’s needs and at the same time provides adequate security. COMSEC engineers test this equipment for reliability in vibration machines and salt-spray chambers. They assure its compatibility with the user’s existing equipment, and they cooperate with the manufacturer to get the best devices at the lowest cost.
In addition to the suggestions that contractors make for improving machines, COMSEC evaluates the hundreds of ideas for new “unbreakable” cipher systems that pour in upon the National Security Agency from amateur cryptographers. The agency gets at least one a day, often channeled to it from the Army or the F.B.I. or the State Department. Many are from professional men, such as doctors and lawyers, but one came in from a prisoner (it was forwarded by his warden). A good percentage include a challenge message, and the COMSEC experts can just visualize the devilish grin of the inventor as he finishes enciphering the message, and thinks, “They’ll never get that!”
The inventors fall into two categories. One type has just read Edgar Allan Poe’s dictum in “The Gold-Bug” that “it may well be doubted whether human ingenuity can construct an enigma of the kind which human ingenuity may not, by proper application, resolve,” and has, in half an hour, invented an unbreakable cipher that disproves it. The other has just devised a cipher so simple that a 12-year-old can operate it (never a 13-year-old), and as a patriotic American is giving it to his government for a mere $100,000—a cheap price for assuring the security of information worth much more than that.
Few of the inventors have any idea of the volume of modern communications, of the conditions under which ciphering is done, of modern cryptanalysis, or that the unbreakable cipher, in the form of the one-time pad, already exists. Nearly all the systems are pencil-and-paper, which are all but useless today, and the chances are almost nil that even a tinkerer in a machine shop will come up with anything new and worthwhile. Nevertheless, COMSEC looks seriously at every proposal. It perhaps recalls that all the basic cryptographic principles now in wide use—the rotor, the Jefferson cylinder-strip system, the one-time tape, the Hagelin mechanism—were created by persons with no professional cryptologic background. The next letter may come from a new Hebern, submitting a valuable concept. Besides, it’s fun to solve the challenge cryptograms—which COMSEC very often does, despite a brevity that would never be met with in practice.
In a way, however, the agency seems to take unfair advantage of these inventors. Their ideas disappear into the black maw of the N.S.A. and may even see service in American cryptography, but security prevents the inventor from ever knowing of this—and may enable the agency or its employees to utilize his ideas without compensation. Fear of this may keep some inventors from submitting potentially valuable ideas. The agency might attract more suggestions by a firm promise not to use ideas without payment; this might be of some value if the matter ever came to court. But the agency will not give such a promise. More incomprehensibly, it will not even say why it will not. It seems that here N.S.A. is being deliberately self-injurious.
COMSEC presides over a great variety of cryptosystems. The Army requires different methods for the differing needs of front-line, middle-echelon, and high-command communications. The Navy’s needs may not vary quite so widely, but even it uses strip cipher for less importan
t communications and rotor machines for more important ones. The Air Force probably uses small codes for its airborne communications and a host of systems for its ground communications, including those to the missile-launch centers.
Are American cryptosystems secure? Different agencies investigate this question in different ways. N.S.A. tests the theoretical limits of security of ciphers. For example, COMSEC mathematicians might calculate the maximum number of messages that could be sent with unchanged primary key (as the wiring in a set of rotors) before enough secondary-key overlaps could be expected to make solution likely. They use such information to prescribe key changes. The individual agencies probably test the practical security of their own systems by monitoring and actual cryptanalysis; the State Department, for example, employs half a dozen cryptanalysts. In addition, independent tests are made, as by the Institute for Defense Analyses. In one case, I.D.A. cryptanalysts were given 1,000,000 letters of error-free text in a top military cryptosystem. They put in the equivalent of six man-years on it—and finally gave up in defeat. The episode speaks well for the security of that cipher, and, by implication, for that of other American cryptosystems.
In a jet-age world, voice communications, with their speed, convenience, personal quality, and two-way nature, are essential. Scramblers keep the conversations private. COMSEC’s functions extend to ciphony, though the Bell Telephone Laboratories also do a great deal of development work.
Scramblers today are vastly improved over the old World War II models, mainly because they employ a new form of telephony called “pulse code modulation.” PCM converts the voice signal into a sequence of pulses and nonpulses, somewhat like a teletype signal. The number of pulses per second varies with voice frequency. This digital form permits the interlacing of many speech signals within a single circuit, thus increasing the capacity of a telephone network. While PCM alone affords some security, since PCM equipment is needed to reconvert it to voice form, its main cryptographic advantage lies in the ease and security of encipherment in the digital mode. The scrambler can encipher the sequence of pulses and spaces just as in the Vernam system. The millions of key pulses can be stored as magnetized spots on metallized tape, as light and dark spots on film, as holes in punched cards. Or a computer can generate them. (A million pulses will last for two and one half minutes of PCM encipherment at 8,000 pulses a second.) Though problems of synchronization afflict PCM systems, they are highly secure, since problems of the voice’s resistance to distortion, which render the continuous-wave scramblers so vulnerable, do not arise.