PRISM-Guardian: Enhancing Data Privacy in Devices With Sound Collection, Recognition, and Sharing Through Blockchain Technology

Abstract

The proliferation of voice-activated devices, such as virtual assistants and voice-controlled systems, has changed how people interact with technology and the environment. These devices collect data that can be sent to servers to process sound, returning responses or suggestions to the user. However, the widespread use of these devices has led to intensive data collection, exposing sensitive information, such as conversations and intimate audio. In this context, we developed PRISM-guardian, a technique for sharing and tracking sound data without revealing its origin, thus preserving privacy. Transparently, audio generators, such as residential users, can track who accessed their information and why. We collected 1000 audio samples, each lasting 10 s, to recognize short-duration cough and sneeze sounds. We achieved average sound recognition processing times of 3.78 s, 6.78 ms to encapsulate the data in the API, and an average of 48 ms to save the data on the blockchain. Besides, we present a mathematical formalization of PRISM and conduct tests to identify the origin of the sound. The results showed that the identity of the sound source is preserved while this source can view and track the data.