Securing Voice Control Technology in Manufacturing with Cross-Domain Authentication

Just as Siri and Alexa have revolutionized the way consumers perform daily tasks, voice-assisted technology is being widely adopted and deployed throughout manufacturing. But the convenience of using voice control technology in factories also opens the door to adversaries seeking to access a company’s critical information and controls. With unauthorized access, attackers can make equipment malfunction, create faulty products, or completely disrupt operations.

Instead of using traditional approaches in the audio domain to defend against attacks, this team will explore capturing voice commands in another domain (e.g., vibration domain) and then perform cross-domain authentication. For example, sensors embedded in wearable devices (e.g., smart watches or activity trackers) will capture users’ speech signals and compare them with audio signals to authenticate the user.

This project will advance the security of voice control technology in manufacturing by using cross-domain sensing capabilities to enhance user authentication. New software will be available to deploy, and new signal processing and machine learning techniques will be developed to study various sensing signals.