Voice assistants and the new security challenge

In September 2019, Amazon announced its Voice Interoperability Initiative: the tech giant wants multiple voice assistants to be accessible through a single device (in the same way that people can use whichever browser they choose to access a website). More than 30 companies are on board with Amazon's initiative, including Microsoft, Spotify, Sony Audio Group and Verizon. However, Amazon's biggest competitors in this market – Google, Apple and Samsung – are not part of the program.

Amazon says its initiative has four main goals: to foster voice services that work seamlessly with each other; to ensure smart devices support multiple wake words; to develop technology that makes it easier for multiple voice services to co-exist on a single device; and to improve the quality of voice interactions through machine learning and conversational AI research.

Voice assistants are an essential component of today's smart homes

Amazon is making the case that we're going to see more households become multi-assistant environments. Customers will ultimately want more flexibility from their devices.

"They don't want to be locked into using a specific voice service," Mariana Zamoszczyk, senior analyst for Smart Living at Ovum, said in a statement provided by Amazon. "This trend means that device makers and AI developers need to prioritize interoperability with other services, and work to deliver differentiated, personalized experiences through their own products or assistants."

Different wake words (a word or phrase that means the voice assistant knows when to start interacting with the user) would be used for different devices, all of which the single speaker would recognize. Multiple simultaneous wake words provide the best option for customers," Amazon CEO Jeff Bezos said in a statement.

Unanswered security questions

Amazon says Initiative participants will work with researchers and universities on a number of challenges: developing algorithms that allow wake words customization; allowing wake words to run on portable, low-power devices; and improving the encryption and APIs that ensure voice recordings are routed securely to the right destination. To achieve this, actual voice recognition (determining the user's identity) will need to be realized quickly.

Amazon is yet to go into detail about how it will safeguard user data. At what point is the individual's data protected? Is it protected as it is being transferred over a network? Is the same level of security required for the 30-plus voice assistants that are taking part? Artificial intelligence providers will need to protect data ownership and ensure that they stay in line with data protection regulations.

While allowing the multiple voice services on the market to be accessible from one device will improve the user experience, there is some way to go in the development of security measures required for the data collected by voice assistants.

Relatec content: Top 5 IoT predictions for 2020