Did These Computer Scientists Solve the Cuban ‘Sonic Attack’?
Three computer scientists reverse-engineered a recording of the sound and think they know what got 24 people at the American embassy in Havana sick.
A technical report from the University of Michigan offers a stunningly simple theory for the source of the Cuban “sonic attack”: a pair of eavesdropping devices too close to each other and tripping the ultrasound that ironically was supposed to make their presence quiet.
More importantly, it might not have been done with malicious intent.
“It doesn’t prove it’s the cause,” Kevin Fu, an associate professor at the University of Michigan and one of the co-authors of the study, cautioned. “It’s a correlation. But to us, it seems like a strong correlation.”
A recap: Last September, the State Department recalled 21 American employees from the U.S. embassy in Havana. These employees, along with three Canadians, reported dizziness, cognitive difficulties, headaches, and hearing loss, among other medical issues, according to an official statement made by Secretary of State Rex Tillerson. The victims of what was being termed a “sonic attack” reported hearing a high-pitched sound that made them physically ill.
In December, the AP reported that patients showed “unprecedented” neurological damage, with hearing loss, memory problems, and cognitive issues. This prompted further speculation as to what the “sonic terror” device was: Some thought it was an advanced Russian tool that was sneaked into the embassy; others thought it was poisoning. A Cuban panel of scientists thought the “psychogenesis” was brought about by stress.
To add to the confusion, doctors weren’t sure either what was going on. In a preliminary report published in JAMA, physicians treating the patients could only say a “novel mechanism” caused the neurological damage. A companion report in JAMA published last month came to no conclusion as to what could possibly be causing the neurological damage patients had suffered.
But in the technical report published from the University of Michigan on Thursday, Fu and his colleagues came up with a totally different, less spy-novella-ish technical mishap, not a “sonic attack” at all.
In fact, Fu and his co-authors were accidental investigators of the Cuban sonic attacks. Fu’s daytime job is researching computer security and privacy at the University of Michigan; he’s also chief scientist at the health-care security company Virta Labs.
“I look at how security can fail,” he told The Daily Beast. “My laboratory studies how sound waves can cause bizarre malfunctions in computer systems.”
About six months ago, the AP released a video of the sound, recorded by a victim at the embassy. (Warning: This sound might be painful to some people.)
Fu and his team were working on a project testing the audibility of ultrasound in another project, but the AP video caught their eye.
“At the time, people were talking about ultrasound [being a theory as to what the sound was],” Fu said. “But it didn’t make sense. Ultrasound is inaudible [to humans], and you wouldn’t hear it.”
So Fu and his co-authors set to work analyzing the five-second blip of sound captured by the AP’s source. Fu and his co-authors decided to “reverse-engineer” the ultrasonic signals necessary to generate the high-pitched tone, attempting to “craft ultrasound with mathematical properties such that you can choose the audible byproduct.”
That’s difficult to parse, but what Fu and his colleagues were trying to figure out were the combination of ultrasonic tones that would not only be able to create something that was discernible to the human ear but would replicate the tinny, high-pitched sound captured by the AP’s source. They knew it wasn’t a single tone, but that there had to be multiple tones that rang together as one. Fu’s team was positive this was the case, because they looked at the spectral signature, or the variance in the wavelengths the tone was emitting.
What they measured came out to be a 7 kHz tone that can be listened to here: (Warning: This sound might be painful to some listeners.)
“That’s 7,000 vibrations a second,” Fu said. “It’s high-pitched, and it’s a sound that any adult or child can hear.”
Fu and his team had figured out how an ultrasound tone could be audible to the human ear. But what could possibly have created that sound?
While ultrasound might make you think of pregnant women, it's in our everyday lives. “The common use of ultrasound is for motion detection,” Fu said, particularly in industrial settings. Those lights that stay on in office rooms? They’re usually able to detect movement (much like the detection of a fetus and its minute movements) in energy-efficient structures to keep the lights on only when there is a moving, living, breathing human in it. Sit still long enough and the lights might shut off, mistakenly thinking that no one is in the room. These sensors are also emitting sound, but it’s at the ultrasound level, in the 32 kHz range, which is normally outside of human earshot.
In an article Fu and one of his co-authors wrote for The Conversation, the authors point to other ways ultrasound can turn up in an industrial setting: museum recordings and security settings that are intended to not bother those outside a setting but respond to those within one; electronic pest repellants that don’t affect humans but annoy rodents and/or bugs; noisemakers designed to affect teens with better hearing than adults in the event of a riot.
But what if the ultrasound here got tripped up by an interruption—perhaps a pair of eavesdropping devices whose transmission got tangled over what was supposed to be an inaudible ultrasonic link but instead became audible?
Fu and his colleagues tested this theory by having an eavesdropping device record conversations that were then sent over to a surveillance team via ultrasonic link, which was supposed to be inaudible to the human ear. But Fu’s group also dropped another otherwise-inaudible ultrasonic device in the vicinity of the first device, creating interference—what’s known as “intermodulation distortion”—that could lead to the 7 kHz tinny sound the team replicated and identified in the AP’s sound recording.
“It doesn’t prove that this is what happened in Cuba,” Fu cautioned. “But it does show that there’s a reasonable probability that it’s an accident rather than someone causing harm [intentionally].”
The technical paper Fu and his co-authors published is groundbreaking in that it offers a viable explanation for what happened to the 24 embassy workers. “It’s an alternative hypothesis to the sonic-weapon theory, of someone trying to cause harm,” Fu said. “It’s a theory that seems a little more practical in that it could be bad engineering.
“It seems like a reasonable hypothesis.”
So many parts of the story seemed primed for a Hollywood thriller: that the embassy workers got sick by virtue of a sound, that the sound seemed targeted at certain times, that only some of them were able to hear it while others weren’t.
Fu said the dual-eavesdropping-device theory could readily explain these oddities. For one, “There’s very little consensus on whether airborne ultrasound can cause harm,” Fu pointed out, saying some research indicates that the answer is yes, others say no. He also said that because ultrasound isn’t usually audible, standards for how loud it can be before it causes harm vary by country. In Canada, for example, the accepted level is 110 dB, the equivalent of “putting your head next to a chainsaw,” Fu said.
But to Fu, “the most interesting part and nuance is that ultrasound can create the audible byproducts and even lower frequencies than can be heard.” That can in turn lead to unusual neurological symptoms—“headaches, dizziness, disorientation.” Sound familiar?
The JAMA study from last month theorized ultrasound was potentially a culprit, noting:
Ultrasound (>20 000 Hz)—specifically high-intensity focused ultrasound—is known to induce heating and coagulative necrosis of brain tissue. This characteristic has recently been exploited to stereotactically and noninvasively produce focal lesions in the treatment of movement disorders. However, the technical challenges in using ultrasound waves for nonlethal attacks include the rapid absorption of ultrasound by surrounding air and a requirement for close proximity to the source to induce injury.
(The Daily Beast reached out to the authors of the companion JAMA report that described the symptoms the patients faced and the fact that they seemed to suffer neurological damage. The authors declined to comment.)
As for the argument that some people were able to hear the sound but others were not, Fu also provides an explanation of varying auditory capabilities, most likely traceable to the simple demographic factor of age. As we get older, our hearing deteriorates. Fu offered a story of doing an experiment one day while playing a couple ultrasonic tones. “A couple students down the way said, ‘Please turn the annoying sound off,’” he recalled. Fu had no idea what they were saying, until he looked at a device he had on that indicated that soundwaves were being generated at the 15 kHz level. “Everyone said they could hear it and it was really annoying,” Fu, who is 42 and reported normal age-related hearing loss, said. “But I couldn’t hear a thing.”
Fu and his colleagues submitted the technical report to the Department of State “some days ago,” but hasn’t heard back. He said that while it’s not probably going to be submitted to a journal because it would be difficult to peer review (the field is tiny and Fu said the experts capable of peer reviewing the report are limited to him and his co-authors, thereby making a peer review moot), the hypothesis makes sense to him.
“It just seems like the simplest solution,” he said.