Inaudible acoustics: Techniques and applications

This dissertation is focused on developing a sub-area of acoustics that we call inaudible acoustics. We have developed two core capabilities, (1) BackDoor and (2) Ripple, and demonstrated their use in various mobile and IoT applications. In BackDoor, we synthesize ultrasound signals that are inaudible to humans yet naturally recordable by all microphones. Importantly, the microphone does not require any modification, enabling billions of microphone-enabled devices, including phones, laptops, voice assistants, and IoT devices, to leverage the capability. Example applications include acoustic data beacons, acoustic watermarking, and spy-microphone jamming. In Ripple, we develop modulation and sensing techniques for vibratory signals that traverse through solid surfaces, enabling a new form of secure proximal communication. Applications of the vibratory communication system include on-body communication through imperceptible physical vibrations and device-device secure data transfer through physical contacts. Our prototypes include an inaudible jammer that secures private conversations from electronic eavesdropping, acoustic beacons for location-based information sharing, and vibratory communication in a smart-ring sending password through a finger touch. Our research also uncovers new security threats to acoustic devices. While simple abuse of inaudible jammer can disable hearing aids and cell phones, our work shows that voice interfaces, such as Amazon Echo, Google Home, Siri, etc., can be compromised through carefully designed inaudible voice commands. The contributions of this dissertation can be summarized in three primitives: (1) exploiting inherent hardware nonlinearity for sensing out-of-band signals, (2) developing the vibratory communication system for secure touch-based data exchange, and (3) structured information reconstruction from noisy acoustic signals. In developing these primitives, we draw from principles in wireless networking, digital communications, signal processing, and embedded design and translate them to completely functional systems