[Paper Review] Privately Connecting Mobility to Infectious Diseases via Applied Cryptography

Alexandros Bampoulidis, A. Bruni+4 more|arXiv (Cornell University)|May 5, 2020

Opportunistic and Delay-Tolerant Networks4 citations

TL;DR

This paper proposes a privacy-preserving protocol that enables secure, private aggregation of mobile phone location data from infected individuals to improve infectious disease modeling without exposing individual identities or infection status. It combines homomorphic encryption, zero-knowledge-inspired masking, and differential privacy to ensure that only a differentially private heatmap of aggregated mobility is revealed, achieving 59-bit statistical security and processing 8 million subscribers in 70 minutes.

ABSTRACT

Recent work has shown that cell phone mobility data has the unique potential to create accurate models for human mobility and consequently the spread of infected diseases. While prior studies have exclusively relied on a mobile network operator's subscribers' aggregated data in modelling disease dynamics, it may be preferable to contemplate aggregated mobility data of infected individuals only. Clearly, naively linking mobile phone data with health records would violate privacy by either allowing to track mobility patterns of infected individuals, leak information on who is infected, or both. This work aims to develop a solution that reports the aggregated mobile phone location data of infected individuals while still maintaining compliance with privacy expectations. To achieve privacy, we use homomorphic encryption, validation techniques derived from zero-knowledge proofs, and differential privacy. Our protocol's open-source implementation can process eight million subscribers in 70 minutes.

Motivation & Objective

• To enable accurate modeling of infectious disease spread by aggregating mobility data from infected individuals only, improving predictive performance in low-prevalence scenarios.
• To prevent leakage of individual location data or infection status during the aggregation process.
• To ensure that neither the health authority nor the mobile network operator can learn more than the final differentially private heatmap.
• To maintain statistical security against malicious parties attempting to infer private data through cheating.
• To provide a practical, open-source implementation suitable for real-world deployment in public health surveillance.

Proposed method

• Uses homomorphic encryption to allow the health authority to securely compute a linear combination of the mobile network operator’s location data using encrypted identifiers of infected individuals.
• Employs a double-masking technique based on random linear combinations of masks to achieve 59-bit statistical security against cheating, reducing the chance of successful forgery.
• Applies a final random mask vector to the output to prevent leakage if any individual output values are known in advance.
• Integrates differential privacy by adding Laplace noise to the final heatmap, ensuring that the presence or absence of any single infected individual cannot be statistically distinguished.
• Combines all privacy mechanisms into a two-party secure computation protocol where the health authority inputs patient IDs and the MNO inputs location data, with no party learning the other’s input.
• Uses a secure evaluation protocol over a finite field with a 60-bit modulus to balance performance and security, with parameters tuned for real-world use in nation-states.

Experimental results

Research questions

• RQ1 Can we securely compute a heatmap of aggregated mobility data from infected individuals without revealing individual locations or infection status?
• RQ2 How can we prevent a malicious party from learning private data through input manipulation or output analysis?
• RQ3 What level of statistical and differential privacy can be achieved while maintaining the utility of the resulting heatmap for epidemiological modeling?
• RQ4 Can the protocol be efficiently implemented at scale, processing large datasets like 8 million subscribers in a reasonable time?
• RQ5 How can we balance privacy guarantees with the accuracy of the final heatmap in low-prevalence epidemic scenarios?

Key findings

• The protocol successfully computes a differentially private heatmap of infected individuals' mobility using secure computation, ensuring that no party learns more than the final output.
• The use of double masking increases statistical security to 59 bits, significantly reducing the probability of successful cheating.
• The open-source implementation processes 8 million subscribers in 70 minutes, demonstrating practical feasibility for real-world deployment.
• The integration of differential privacy ensures that individual locations cannot be re-identified, even if an attacker has auxiliary knowledge about the data.
• The protocol maintains high utility of the heatmap for epidemiological modeling, especially in low-prevalence scenarios where population-wide mobility data may be less accurate.
• The system is resilient to both honest-but-curious and malicious behavior from either party, as proven in the security analysis.