Accurate crowd statistics are essential for urban safety, smart city services, and event management, but also risk exposing sensitive information. Precise counts can enable adversaries to identify individuals by correlating released statistics with external data. We propose a novel protocol for privacy-preserving crowd analytics that combines Bloom filters, fully homomorphic encryption (FHE), and differential privacy (DP). Unlike prior approaches without formal guarantees, our design injects randomized DP noise directly under encryption, providing strong and quantifiable privacy protection. All computations, including aggregation and noise addition, are executed entirely in the encrypted domain, ensuring that no intermediate values are leaked. Our evaluation shows that the protocol achieves both efficiency and accuracy: relative accuracy exceeds 90% with 𝜖 = 0.5, and aggregation of over 1000 users completes in under one second on commodity hardware. By uniting encrypted processing, formal privacy guarantees, and practical scalability, this work demonstrates the feasibility of deploying privacy-preserving crowd monitoring in real-world smart city infrastructures.
Authors: Fatemeh Marzani et al
Publication date: 2026