A Layered Privacy Architecture for Utility-Preserving Internet of Things Systems

Abstract

Internet of Things (IoT) ecosystems face intense pressure to balance data utility with user privacy, particularly as sensor deployments scale into millions of endpoints. Existing work on IoT privacy preservation techniques (PPTs) spans cryptographic PETs, differential privacy, and AI-assisted data minimization strategies, yet integration into real-world architectures remains challenging due to heterogeneity, resource constraints, and deployment diversity. This paper proposes a layered privacy architecture that embeds lightweight secure computation, decentralized identity, and adaptive anonymization gates within the IoT data pipeline. The architecture was prototyped on a smart city testbed consisting of over 3,200 heterogeneous sensors and gateway nodes. Results demonstrate a substantial reduction in unintentional data leakage, measured by mutual information metrics, while maintaining 82–90% of analytic utility in common monitoring tasks. The proposed design offers a practical roadmap for practitioners seeking to deploy privacy-aware IoT infrastructures without onerous overhead or centralized trust assumptions.