The initial idea was that the main challenge facing the South Vietnamese army in fighting the Communist insurgents on its own territory was simply finding the enemy: it was, ASA analysts thought, a problem much like the Navy faced in locating an enemy naval force amid the vast expanses of the ocean, and the solution was thus the same as well. Direction-finding (DF) units would generate a fix on enemy units when they used their radios; South Vietnamese forces would then be dispatched to root them out.
They had not, however, reckoned with what it was like to operate DF equipment in a heavy jungle. The moist air and heavy foliage so attenuated HF radio signals that the ground wave, traveling directly from a transmitter, could not be detected more than ten or fifteen miles away. Some of the Communist transmitters used as little as 1 watt of power, making detection even more difficult. The sky waves produced when HF signals bounced off the atmosphere traveled longer distances, but came down at such a sharp angle that they could not be detected with the old equipment the South Vietnamese army had inherited from the French. In any case, there was a ninety-mile-wide “skip zone” where neither the direct ground wave nor the sky wave could be heard at all, effectively blacking out most of South Vietnam from any listening post based in Saigon. The only solution was to put DF equipment in mobile units, consisting of a three-quarter-ton truck and two jeeps.
The result, in the words of an internal NSA history of Vietnam, was “sheer chaos.” The field missions had to get dangerously close to their targets and even then could detect only about 5 percent of enemy transmitters. The danger was made all too clear when a DF unit was ambushed ten miles from Saigon while returning from a mission on the south coast on December 22, 1961; an ASA soldier, Specialist James T. Davis, and nine South Vietnamese were killed. Davis was the first American fatality of the Vietnam War.20
The ambush hastened a project that the Army already had under way which it thought could solve both the technical and logistical problems of obtaining radio fixes on the Communist forces: namely, to mount DF equipment on light aircraft. By March 1962, Army engineers had overcome a puzzling technical challenge to HF airborne direction finding and had a system flying in Vietnam on a single-engine plane, the U-6A Beaver, and working well. (The system basically used the plane’s metal skin as an antenna, turning what had seemed an insoluble interference problem into a virtue; instead of rotating a movable loop antenna to determine at what direction the signal was the strongest, the pilot yawed the plane itself in a fishtailing course until a pair of fixed antennas mounted on each wing registered the same intensity. He would then fly to two different spots and obtain additional bearings to complete the triangulation of the target.)21
The American SIGINT specialists had not reckoned, however, with how South Vietnamese or American military commanders would use the information obtained by such technological innovation. Under orders from South Vietnam’s unelected strongman, Ngo Dinh Diem, who feared that his troops would not be available to protect him from coups in Saigon, his army was forbidden to engage in actions that might result in casualties to themselves; when they were alerted to the presence of Communist forces, the South Vietnamese moved slowly or not at all, inching forward in American-supplied armored personnel carriers only after bombing and strafing indiscriminately with close support aircraft, thereby invariably giving the insurgents time to get away unscathed.
A failed South Vietnamese army assault at Ap Bac, south of Saigon, in January 1963, launched on the basis of an accurate DF fix on a Communist transmitter, was a more serious debacle: the Communist forces shot down five helicopters, damaged nine others, and inflicted two hundred South Vietnamese casualties for the loss of a handful of their own fighters. General Harkins hailed the capture of the village as a decisive victory. But Ap Bac would prove a turning point in American military involvement in Vietnam, convincing U.S. commanders that American ground troops would ultimately have to carry the burden of the fight.22
Yet throughout the war U.S. troops were often no better at employing DF fixes. The Communists quickly learned to place their radio huts some distance from command posts, and not infrequently U.S. ground commanders called in strikes to “blast a patch of jungle just because a transmitter had been heard there,” as NSA’s declassified history related. Fundamentally, the problem was that a generation of U.S. Army commanders had grown up without learning to put signals intelligence to practical use on the battlefield the way their World War II predecessors all had. “Very few commanders had any training in SIGINT. In the 1950s it had been kept closeted, a strategic resource suitable only for following such esoteric problems as Soviet nuclear weapons development.” Later in the war when airborne intercept posts on RC-130 aircraft were able to instantly spot and report activations of SA-2 surface-to-air missile radars in North Vietnam, the Air Force refused to act on the information because SAC procedures called for photoreconnaissance confirmation before carrying out any strikes on air defense sites, another instance of Cold War doctrine trumping battlefield realities.23
Adding to the difficulties, all of the old fights over control of signals intelligence in the field resurfaced. The hard-won lessons from previous wars of the importance of centralization seemed to have been utterly forgotten; it was as if Korea or World War II had never happened. The Army and the Military Assistance Command Vietnam furiously opposed NSA’s move to take charge of ASA’s field center, which had relocated from Tan Son Nhut to Phu Bai in 1962, protesting any attempt at “removing these SIGINT resources from the control of military commanders in the area.” NSA’s director, Gordon Blake, subsequently worked out a deal directly with the Joint Chiefs of Staff under which NSA would run all fixed sites within the country while tactfully agreeing to “delegate” its authority to ASA to command “direct support units” that operated with U.S. troops, effectively placing them under the control of local Army commanders.
The result, as usual of such compromises, tried to satisfy everyone while leaving a mess that only exacerbated suspicions and left NSA and the service cryptologic agencies duplicating and tripping over one another. The Air Force, which wanted its own airborne DF capacity, tried and failed to develop a workable system early on but in 1966 was back with four dozen specially equipped C-47s—and insisting that because they were really doing ELINT, not COMINT, the operation came under Air Force rather than NSA control.24 NSA’s unilateral move to shift signals intelligence processing from the field to a center at Clark Air Base in the Philippines, and later Okinawa and then finally Fort Meade, aroused suspicions of local commanders that NSA was not giving its full support. So did NSA’s abrupt decision, following a series of communications changes by Communist forces in April 1962, to essentially abandon efforts at cryptanalysis of high-level systems altogether and focus on DF and traffic analysis instead.
The enemy code changes—which NSA officials more out of reflexive habit than sense immediately blamed on a recent leak by South Vietnamese government sources of U.S. cryptanalytic success—were “catastrophic for the American SIGINT effort,” in the words of NSA’s Vietnam history. In fact, the move had clearly been in the works for some time, and could not have been a reaction to recent events, as it involved not only new procedures for frequent and regular changes in call signs, schedules, and radio frequencies but also entirely new codebooks and additive pads, which would have taken months to prepare and distribute. Throughout the war U.S. cryptologists would successfully read low-and medium-grade codes employed at the tactical level by enemy units at the level of regiment and below (so much so that by 1968 the processing and decryption of this traffic would have to be automated to handle the considerable volume), but tackling Hanoi’s higher-level codes was considered unlikely to be worth the investment of time NSA thought they required.25
One of the other forgotten lessons about signals intelligence in a real war would prove even more costly to U.S. forces. In 1965, NSA analysts examining traffic on a network used by Chinese air forces in Vietnam noticed that some of the messages
began with an unusual character, a “barred E,” . .—. . in Morse code. Recalling that this same character appeared as a prefix on urgent messages transmitted by German U-boats in World War II to report convoy sightings, the chief of B21 division, E. Leigh Sawyer, suggested on a hunch that they compare the timing of these messages with the launch of U.S. bombing missions against North Vietnam. (Named Operation Rolling Thunder, the airstrikes had begun in March 1965 in an effort to halt the movement of supplies to the South and to increase political pressure on North Vietnam.) The correlation was perfect: barred E messages had been sent ahead of 90 percent of Rolling Thunder strikes that targeted the northeast quadrant of the country. The warnings were giving North Vietnamese MiG pilots time to scramble and be waiting—and add to the toll of more than nine hundred U.S. aircraft shot down during the three years of Rolling Thunder.26
Under a program called Purple Dragon, NSA teams were sent to look for leaks in U.S. communications that might be tipping off the enemy. They found a torrent. Tactical commanders had received as little instruction in communications security as they had in the use of signals intelligence. American radio operators frequently made up their own amateurish codes rather than follow required procedures. Strict rules required commanders to account for all code materials and return them for destruction at regular intervals; rather than risk getting a black mark for having to account for lost items, the NSA investigators found, many simply locked the codes up in their safe rather than distribute them for use in the field. As a test, NSA experts playing the part of an enemy analyst tried to see how much they could learn from intercepted U.S. signals. From traffic analysis of a single encrypted voice channel between two air bases, they were able to successfully predict eighteen of twenty-four actual air missions over the North.
The North Vietnamese army, which had a signals intelligence staff of five thousand, proved as adept at exploiting traffic analysis as NSA was. Every U.S. bombing mission was preceded by an upsurge of traffic involving logistics, ordnance loading, weather flights, and aerial refueling tankers, and even if none of the content of the signals was readable, the pattern was a dead giveaway.27
The U.S. Air Force generally spared the North Vietnamese even that trouble. Reflecting bureaucratic inertia, American overconfidence, and more than a little disdain for the intelligence capabilities of the enemy, nearly all radio communications of the U.S. air operations used unencrypted tactical voice. NSA’s efforts to have the Air Force install voice encryption equipment on aircraft had gone nowhere for years: as one NSA official involved in the program remarked in frustration, the Air Force would accept such a device “only if it had no weight, occupied no space, was free, and added lift to the aircraft.” The Air Force insisted that air operations moved too quickly to require such security measures anyway.
But a trove of North Vietnamese signals intelligence documents subsequently captured revealed that U.S. Air Force plain-language voice had been the North’s major source of advance warning of U.S. airstrikes throughout the war. Between January and September 1966, 228 aircraft were shot down over North Vietnam, and the captured documents showed that the North Vietnamese had at least thirty to forty-five minutes’ warning of 80 to 90 percent of Rolling Thunder missions. Despite NSA’s occasional success in tightening up particularly leaky communication practices, the problems continued throughout the war. SAC, which flew B-52 bombers from Guam to strike Communist forces in the South starting in June 1965 in an operation called Arc Light, was by far the worst offender, giving the North Vietnamese as much as eight hours’ warning and often revealing exact launch times and likely targets.28
It had long been an article of unshakable NSA doctrine that the least disclosure of success in exploiting enemy communications would instantly cause the source to be lost. But the serene indifference of U.S. forces seemed to be proving exactly the opposite proposition: no matter how glaring the evidence that one’s communications systems had been compromised, it was impossible to get anyone to do anything about it.
—
Even as McNamara began to doubt by 1967 that airstrikes could force North Vietnam to halt its operations against the South and accept a negotiated end to the war—the longest sustained bombing campaign in history, Rolling Thunder would deliver 643,000 tons of bombs by the time Johnson ordered a halt in 1968—the defense secretary retained an unwavering confidence that high-tech wizardry held the ultimate solution to defeating an elusive and resourceful enemy in a jungle counterinsurgency campaign.
By the end of 1967 construction began of an elaborate networked system of ground sensors to locate and target infiltrators and supplies moving into the South through rainforest trails in Laos and South Vietnam. McNamara’s Wall was the U.S. answer to North Vietnam’s Ho Chi Minh Trail. It cost billions of dollars by the time it was done. Dropped from aircraft and camouflaged to resemble vegetation, twenty thousand electronic sensors picked up sounds, vibrations of footsteps, magnetic fields from a passing soldier’s rifle, electrical emissions from a nearby truck engine, even human urine. Aircraft orbiting overhead relayed radio signals from the sensors to a data center in Thailand, where IBM 360/65 computers correlated all the information and controllers called in airstrikes on identified targets. NSA contributed to the effort with a series of wiretapping devices for intercepting landlines along the trail routes: early models required a man to stay behind, hidden in the bush, ready to detach the tap if a Communist patrol approached to inspect the line, but later models were designed to be indistinguishable from standard Vietnamese insulators. NSA engineers subsequently devised a helicopter-dropped pole that could pick up signals from a nearby landline without any direct connection at all.29
The other high-tech, computer-based intelligence-processing marvel of the war was an NSA and Air Force system that recreated the tactical warning system for fighter aircraft that had proved so effective in Korea. It was slow in coming: once again the lessons of an earlier war had been largely forgotten and had to be reinvented from scratch. An intercept site at Da Nang Air Base on the coast near the DMZ, was able to pick up the highly stereotyped radar tracking plots sent in manual Morse by North Vietnamese radar sites to air defense headquarters; these consisted of a series of numbers or letters giving the altitude, speed, direction, and identity of aircraft, both friendly and enemy. As in Korea, this was an irreplaceable source of distant warning of incoming MiGs that threatened U.S. aircraft. But the security rules that required concealing the source of signals intelligence proved a maddening obstacle to putting it to use: the intercept stations were forbidden to pass the information on to the Seventh Air Force until the enemy aircraft were within range of a U.S. radar station that could plausibly have been the source of the warning.
Even then, the Air Force insisted on a convoluted procedure intended to ensure that its own commanders called the shots but which added deadly delays. The intercept site at Da Nang passed the information to the Air Force Tactical Air Control Center at Tan Son Nhut Air Base in Saigon, which “validated” the tracks and relayed the information to an Air Force radar center on Monkey Mountain, back near Da Nang, which could then warn aircraft. The whole process took twelve to thirty minutes. Similarly, RC-130 aircraft flying over the Gulf of Tonkin that were able to monitor North Vietnamese air-ground VHF voice channels were not allowed to talk directly to pilots at all. In April 1965 two F-105s were shot down by MiGs even though an RC-130 had information that could have warned them in time of the approaching threat.30
The system that NSA developed to replace this jury-rigged arrangement in 1967 used a computer link that automatically translated into a geographical coordinate the North Vietnamese tracking data entered by a U.S. operator at a manual Morse intercept terminal; those coordinates were then fed directly into the main air defense computer at TACC and were automatically integrated with the plots that showed up on U.S. Air Force controllers’ radar screens. The system, called Iron Horse, reduced the time from intercept to warning to less than three minutes, and so
metimes as little as eight seconds. A later system called Teaball, used to vector F-4s to intercept MiGs that threatened B-52s striking targets in North Vietnam, took voice intercepts of North Vietnamese ground controllers collected by a SIGINT U-2 and passed the information to U.S. air controllers; it was credited with thirteen of nineteen MiGs shot down during its two months of operation during Operation Linebacker, the American bombing of Hanoi and Haiphong ordered by President Richard Nixon in 1972 to try to force the North Vietnamese to the negotiating table.31
—
The scientific aura that always imbued signals intelligence cut two ways. The flip side of a failure to understand and effectively use SIGINT by commanders who were mystified by or distrustful of its esoteric complexities was an overreliance on SIGINT by others who were dazzled by its apparent infallibility. By 1966, NSA analysts had worked out a series of “SIGINT indicators” that reliably indicated a pending Communist military operation; these included new call signs and other unscheduled changes in radio procedures, the activation of networks to communicate with forward command and observation posts, the sudden movement of known transmitters by ten kilometers or more, and an increasing tempo of signaling leading up to the actual moment of attack. But it all depended heavily on interpretation and inference. As accurate as the results obtained through traffic analysis and airborne DF proved to be as the ground war intensified throughout 1966 and 1967, there was no guarantee that enemy intentions were always going to be what analysts had assumed them to be: in the absence of decipherment of high-level messages, it was impossible to know for sure why an enemy unit had moved. And even as larger, regular North Vietnamese and southern Communist units increasingly began to enter the fighting in the South, no action as of late 1967 had ever involved the movement of more than one division at a time, which minimized having to deal with the greatly complicating intelligence problem of identifying diversions staged as part of a coordinated attack on multiple objectives.32
Code Warriors Page 33