In October 2013, the US Coast Guard received a distress call from a purported group of five adults and two children in a boat sinking off the coast of North Carolina.
“The audio analysis and forensics (produced by the Carnegie Mellon researchers) are very impressive. It’s a great partnership we’ve forged here.”
USCG Ensign Gianfranco Palomba, Search & Rescue Hoax Call Project Manager
“Mayday! Mayday! Help! Help! We’re sinking!”
Within minutes, emergency responders, their hearts racing, had deployed several boats and a helicopter in a dash to save the seven from drowning. The caller made repeated pleas for help over the next 50 minutes as responders from the Coast Guard, Marine Corps Air Station, a salvage company, the local Sheriff’s office and two fire departments joined the rescue effort.
The boat was never found.
The call was a hoax.
On June 11, 2015, following a Coast Guard investigation, Brandon Paul Garner, 32, of Beaufort, NC, was sentenced to 28 months of imprisonment and ordered to pay $18,999 in restitution.
Just One of Hundreds
Gianfranco Palomba, an ensign in the USCG’s Research and Development Center, says the Coast Guard receives well over 100 instances of phony distress calls a year—and the number is growing.
Palomba manages the Coast Guard Search and Rescue Hoax Call Project, which is developing and testing advanced technologies to identify and prosecute hoax callers.
These calls are more than a nuisance, and they certainly are no joke.
Every distress call the Coast Guard receives over its VHF radio channel 16—the mariner’s “911”— compels the federal agency to launch an expensive search-and-rescue effort involving at a minimum a small USCG rescue boat, a C-130 fixed-wing aircraft, and several personnel. The cost of each response, Palomba said, can run as high as $200,000 to $250,000.
Not only do phony distress calls cost taxpayers in wasted rescue-operations funding, but they also endanger responders and divert USCG manpower and equipment from critical missions, such as drug interdiction and enforcement of maritime laws. “We see a phenomenon that’s not going away,” Palomba said. “In fact, it’s steadily increasing every year.”
Hoax distress calls are a federal felony carrying a maximum penalty of up to 10 years in prison, a $250,000 fine, a $5,000 civil penalty and possible restitution to the Coast Guard for search and rescue costs.
Unfortunately, these penalties have not been a sufficient deterrent, perhaps because in the past too few hoax callers have been caught.
The Coast Guard hopes to change that.
Experts in Voice Forensics
The only evidence hoax callers leave behind, at least initially, is a raw VHF radio recording of their voice and an initial line of bearing between the origin of transmission and one of hundreds of USCG Rescue 21 radio towers that dot the US coastline.
The voice recording does not reveal the identity of the caller. However, it can reveal a great deal about what kind of person they are, such as their age, gender and ethnicity—information that with other evidence can help win a successful prosecution.
The process of gleaning this information is called voice forensics, and it’s one of the tools being evaluated by the Coast Guard for its Search and Rescue Hoax Call Project.
The Coast Guard’s voice-forensics partner in the project is a team of research scientists at Carnegie Mellon University’s Language Technologies Institute. The team is led by Rita Singh, a senior systems scientist at Carnegie Mellon’s School of Computer Science and a published expert in voice forensics.
The Carnegie Mellon team is part of a homeland-security research consortium called CCICADA, which is funded by the US Department of Homeland Security (DHS). The Coast Guard initially contacted CCICADA Director Fred Roberts at Rutgers University for help on the hoax call project, having worked with him on other USCG security projects.
The theory behind the Carnegie Mellon team’s work is this: Much like a fingerprint, the “voice print” of every person theoretically has unique characteristics.
To produce a voice-print profile of an individual, the Carnegie Mellon team uses an approach called micro-articulometry. This requires the precise measurement and analysis of the micro properties of “movements, dimensions and positions of the articulators in the human vocal tract during the process of speech production.”
Like Bar Codes
Researchers break the recorded voice into tiny segments, and then employ advanced algorithms and machine learning to analyze the movement of sound through each segment. What they end up with, says Singh, is something akin to a unique bar code.
Even when hoax callers distort their voices, which most do, they ultimately produce the same voice-articulation patterns as they would in normal speech.
“You cannot change how fast your muscles move, your vocal tract inertia, your lung volume, your skeletal structure, and all that influences voice,” Singh explained in a recent interview with Information Week. “It’s still ongoing research, but we are discovering things about how we can home in on those aspects of voice and speech that will allow us to make reliable and accurate estimations from the voice about the person’s characteristics.”
Palomba said this kind of analysis can produce not only bio-metric detail about the caller, but also clues about the physical location from which the call was made. For example, he said, it’s possible to determine, even from a poor-quality recording, that a 35-year-old Caucasian male of a certain height and weight made a distress call from a small room with concrete walls and a fan humming in the background.
The Coast Guard has not publicly announced any prosecutions in which Carnegie Mellon’s voice-forensics work has played a role, but its data now figures in several ongoing investigations, Palomba said.
He noted that the research has entered the test and evaluation stage, which, if successful, could lead to more widespread use of voice forensics by the Coast Guard. This would give Commanding Officers the data they need to quickly launch investigations of hoaxers who make serial distress calls.
It would also produce evidence that could be used in well-publicized criminal prosecutions to deter would-be hoaxers.
“The Coast Guard’s goal is to use successful prosecutions as a large-scale deterrent,” Palomba said.
Looking back over his work on this project, he said one thing is now perfectly clear about hoax callers.
“What we have found more and more is that this is not your respected mariner who is exploiting Channel 16. It’s someone who does not have respect for the Coast Guard’s search and rescue mission” such as criminals seeking to divert attention or a miscreant man making calls from a VHF radio in his mother’s basement.
The Coast Guard hopes to continue its relationship with the Carnegie Mellon research team and other DHS-supported university research centers to make further advances in voice forensics and other technology tools critical to its mission.
“The audio analysis and forensics (produced by the Carnegie Mellon researchers) are very impressive,” he said. “It’s a great partnership we’ve forged here.”