Privacy Preserving Event Detection

TL;DR

This work addresses privacy-preserving event detection in a sensor network by formulating a generalized K-sample problem using the marginal distributions of sensor measurements. It introduces a diameter-based statistic, specifically the Hellinger diameter, and shows that using marginal types suffices to achieve the best possible type-I error exponent under a vanishing type-II requirement. To protect sensor data, the authors propose zero-modulo-sum (ZMS) obfuscation implemented through a CPA-secure public-key protocol, enabling the fusion center to compute the diameter-based statistic without revealing individual sensor data. They provide a cryptographic attacker–challenger analysis proving that any PPT adversary cannot significantly improve its probability of inferring sensors’ marginal types beyond independent guessing, as long as CPA security holds. The approach is validated via a simulation in a spectrum sensing scenario, demonstrating practical feasibility and robustness to privacy constraints, with the key insight that privacy protection does not degrade the asymptotic detection performance under the proposed framework.

Abstract

This paper presents a privacy-preserving event detection scheme based on measurements made by a network of sensors. A diameter-like decision statistic made up of the marginal types of the measurements observed by the sensors is employed. The proposed detection scheme can achieve the best type-I error exponent as the type-II error rate is required to be negligible. Detection performance with finite-length observations is also demonstrated through a simulation example of spectrum sensing. Privacy protection is achieved by obfuscating the sensors' marginal types with random zero-modulo-sum numbers that are generated and distributed via the exchange of encrypted messages among the sensors. The privacy-preserving performance against "honest but curious" adversaries, including colluding sensors, the fusion center, and external eavesdroppers, is analyzed through a series of cryptographic games. It is shown that the probability that any probabilistic polynomial time adversary successfully estimates the sensors' measured types cannot be much better than independent guessing, when there are at least two non-colluding sensors.

Figures (9)

• Figure 1: The CPA experiment.
• Figure 2: The source and sensor regions in the crowd spectrum sensing example.
• Figure 3: Plots of $-\frac{1}{t} \log_2\mu_t$ vs. $t$ for different bounds on $\lambda_t$.
• Figure 4: The ROC curves for the networks with 7 and 8 sensors.
• Figure 5: The TEA experiment.

Theorems & Definitions (13)

• Definition 1
• Definition 2
• Definition 5
• Theorem 6
• proof
• Theorem 7
• Lemma 8
• Lemma 9
• Lemma 10
• Lemma 11