The Data61 team at the Commonwealth protect users Scientific from voice spoofing attacks. and Industrial Research Organisation (CSIRO) , in Australia , together with Samsung Research and Sungkyunkwan University in South Korea , have designed and developed a solution to
This solution is called “Voice liveness detection” (Void) and as its name suggests, it is designed to detect whether the voice being heard is actually coming from a human. It is designed to be integrated into a smartphone or voice assistant software and is able to detect the difference between a “live” human voice and a voice played through a speaker, a tactic often followed by hackers to impersonate someone else and fool systems.
According to Data61, Void uses deep learning models to detect voice spoofing and relies on information from spectrograms, a visual representation of the frequency spectrum, to detect whether the voice is indeed “live.”
CSIRO Data61 researcher Muhammad Ejaz Ahmedexplained how spoofing attacks are becoming more common as voice technologies are now used to shop online, make phone calls, send messages, control smart home devices and access banking services.
"Although voice spoofing is known as one of the easiest attacks, as it simply involves recording the victim's voice, it is incredibly difficult to detect because the recorded voice has similar characteristics to the victim's real voice," he said.
“Void is a game-changing technology that enables more efficient and accurate detection, helping to prevent misuse of people’s voice commands.”.
Void was developed through extensive testing. According to Data61, the new technique was able to accurately detect almost all frauds. It also detected attacks eight times faster than other deep learning techniques.
The results from the tests have been published in a research paper titled, “Void: A fast and light voice liveness detection system,” which will be presented at a conference in August.
