Detection and Localization of Acoustic Vulnerabilities of Underwater Data Centers for Remote Surveillance
Conference
Blow, D, Abdullah, A, Sheldon, J et al. (2025). Detection and Localization of Acoustic Vulnerabilities of Underwater Data Centers for Remote Surveillance
. SMART BIOMEDICAL AND PHYSIOLOGICAL SENSOR TECHNOLOGY XI, 13482 10.1117/12.3053819
Blow, D, Abdullah, A, Sheldon, J et al. (2025). Detection and Localization of Acoustic Vulnerabilities of Underwater Data Centers for Remote Surveillance
. SMART BIOMEDICAL AND PHYSIOLOGICAL SENSOR TECHNOLOGY XI, 13482 10.1117/12.3053819
Underwater Data Centers (UDC) offer natural cooling and built-in security for critical naval infrastructures. However, UDC storage devices are vulnerable to acoustic injection attacks as sound travels four times faster in water. In this work, we propose a subsea surveillance framework to identify and interpret acoustic attacks on UDCs in real-time. Our framework includes the system design of position error signal (PES)-based energy signature modeling, an integrated COMSOL simulation engine for comprehensive attack rehearsal, and a data-driven algorithm for spatial-temporal-frequency domain analyses to map such attacks. Specifically, we investigate various acoustic attack conditions on a real UDC testbed as well as in COMSOL simulation for PES-based energy signatures for comprehensive data-driven modeling. We demonstrate that classical approaches such as a K-means clustering model can then be trained to detect potential attacks in real-time from the energy signatures inside UDC pods. We are also in the process of deploying a full-size UDC pod to the ocean for more comprehensive field experimental validation. The subsea pod will ensure real operational scenarios and open up promising opportunities for future research on UDC surveillance.
