The idea of voice activated commands is appealing and systems like Google Home and Amazon Echo are potentially the mere beginnings of a series of technological functions that will be controlled by human voice instructions. The premise is simple – by removing the need to use buttons, dials and switches, consumers can easily operate appliances with their hands full or while doing other tasks. Providing a glimpse of this emerging future, the website Hongkiat provides 15 new types of technology that run on voice commands.
Smart homes and smart cars
Voice activated technology is already being used to access a range of services. This includes financial products; for instance Barclays bank has started to pilot technology that uses voice recognition to verify the identity of call-in center customers. While such technology remains in its infancy, it’s easy to speculate about the possibilities for the future smart home and smart cars. Self-driving cars in particular are set to have a plethora of voice activated devices, perhaps centralized in the form of a software that resembles Siri, but for cars. In parallel with improvements to technology, security also needs to be tightened.
Security risks
Voice activated systems are vulnerable to hacking. This is because sound is regarded as an “open channel”; and this inlet can be be easily spoofed by impersonators. In many cases the impersonation does not have to be particularly good for while the technology that does things when instructed to by voice is getting better, the actual software that recognizes the voice remains less sophisticated and hence more open to being ‘tricked’. Moreover, it is also relatively simple for a more equipped hacker to record someone’s voice and then use a voice recording to access a security system. This all makes for the voice biometric to be a generally weak form of cybersecurity.
How might the security risks associated with voice activated devices be overcome? The approach taken by the University of Michigan researchers comes in the form of a security-token necklace, ear buds or eyeglasses. This approach is based on having a second channel to authenticate the owner of the voice.
For this the researchers developed VAuth. This is a wearable device that continuously registers speech-induced vibrations on the user’s body. These sound waves are then paired up with the sound of that person’s voice, and this conjugation creates a unique and secure signature. The wearable device (accelerometer) that assesses the sound vibrations is in the form of the necklace, ear buds or as a small attachment to eyeglasses.
Protoype
To test out how the VAuth might work in practice, the U.S. scientists developed a prototype based on a standard accelerometer that measures motion, together with a Bluetooth transmitter. The Bluetooth component transmits the vibration signal to the microphone housed in the user’s chosen device.
For the study, the researchers tested out VAuth with eighteen users and thirty voice commands. With these experiments VAuth achieved a 97 percent detection accuracy plus a less than 0.1 percent false positive rate. This is captured in a white paper titled “Continuous Authentication for Voice Assistants.”
The video below also summarizes the study:
The research was also presented to International Conference on Mobile Computing and Networking, MobiCom 2017, in Snowbird, Utah, which took place in October.
Essential Science
This article is part of Digital Journal’s regular Essential Science columns. Each week Tim Sandle explores a topical and important scientific issue. Last week we looked at how gold nanoparticles have been deployed for the detection of a rage of blood related disorders. The week before, we considered a significant advancement has been made with a medical device, in the form of a dialysis membrane, with the ‘super-material’ graphene at its core.
