Oblivious Sensor Fusion via Secure Multi-Party Combinatorial Filter Evaluation
MetadataShow full item record
This thesis examines the problem of fusing data from several sensors, potentially distributed throughout an environment, in order to consolidate readings into a single coherent view. We consider the setting when sensor units do not wish others to know their specific sensor streams. Standard methods for handling this fusion make no guarantees about what a curious observer may learn. Motivated by applications where data sources may only choose to participate if given privacy guarantees, we introduce a fusion approach that limits what can be inferred. Our approach is to form an aggregate stream, oblivious to the underlying sensor data, and to evaluate a combinatorial filter on that stream. This is achieved via secure multi-party computational techniques built on cryptographic primitives, which we extend and apply to the problem of fusing discrete sensor signals. We prove that the extensions preserve security under the semi- honest adversary model. Though the approach enables several applications of potential interest, we specifically consider a target tracking case study as a running example. Finally, we also report on a basic, proof-of-concept implementation, demonstrating that it can operate in practice; which we report and analyze the (empirical) running times for components in the architecture, suggesting directions for future improvement.
Curran, William E (2021). Oblivious Sensor Fusion via Secure Multi-Party Combinatorial Filter Evaluation. Undergraduate Research Scholars Program. Available electronically from