A Cloud-Based Secure and Privacy-Preserving Clustering Analysis of Infectious Disease

Jianqing Liu, Yaodan Hu, Hao Yue, Yanmin Gong, Yuguang FangPAC 2018 · OpenReview

Abstract

The early detection of where and when fatal infectious diseases outbreak is of critical importance to the public health. To effectively detect, analyze and then intervene the spread of diseases, people's health status along with their location information should be timely collected. However, the conventional practices are via surveys or field health workers, which are highly costly and pose serious privacy threats to participants. In this paper, we for the first time propose to exploit the ubiquitous cloud services to collect users' multi-dimensional data in a secure and privacy-preserving manner and to enable the analysis of infectious disease. Specifically, we target at the spatial clustering analysis using Kulldorf scan statistic and propose a key-oblivious inner product encryption (KOIPE) mechanism to ensure that the untrusted entity only obtains the statistic instead of individual's data. Furthermore, we design an anonymous and sybil-resilient approach to protect the data collection process from double registration attacks and meanwhile preserve participant's privacy against untrusted cloud servers. A rigorous and comprehensive security analysis is given to validate our design, and we also conduct extensive simulations based on real-life datasets to demonstrate the performance of our scheme in terms of communication and computing overhead.

Introduction

Early detection is the "holy grail" of public health intervention. To stop an outbreak, epidemiologists need to know exactly where and when people are getting sick. Traditionally, this data is collected via manual surveys or field workers—processes that are slow, expensive, and often suffer from low participation rates because people are wary of sharing sensitive health and location history.

While modern participatory systems exist, they face three major hurdles: privacy concerns regarding exact movements and symptoms, data integrity issues from potential Sybil attacks (fake data submission), and the difficulty of building dedicated global infrastructure. This paper proposes a framework to leverage ubiquitous cloud services like Google (for location) or health-tracking apps to collect data while strictly maintaining user privacy.

System Architecture

The proposed framework introduces a multi-entity system involving a Trusted Authority (TA), a Public Health Office (PHO), and two separate cloud providers: a Location Cloud (LC) and a Health Cloud (HC). This separation of duties ensures that no single entity has a complete view of a user's identity, location, and health status.

System architecture showing the interaction between the Service Plane (Clouds and Public Health Office) and the User Plane.

The protocol operates in three distinct phases:

  1. Anonymous & Sybil-Resilient Authentication: Using Group Signatures, users prove they are valid participants without revealing their identity. A specific check detects and blocks "double-registration" within the same time window to prevent Sybil attacks.
  2. Privacy-Preserving Data Collection: Users send location data to the LC and health data to the HC using Identity-Based Encryption (IBE). Because the data is split across two clouds and encrypted with pseudonyms, neither cloud can link a specific person’s location to their health status.
  3. Key-Oblivious Inner Product Encryption (KOIPE): This core technical contribution allows the PHO to perform spatial clustering analysis without accessing raw data.

Privacy-Preserving Spatial Analysis

To find geographical "hotspots," the PHO utilizes the Kulldorff scan statistic, which requires the sum of infected people in specific areas. The KOIPE mechanism allows the PHO and the Health Cloud to collaboratively compute these sums using Paillier homomorphic encryption.

Protocol sequence for the Key-Oblivious Inner Product Encryption (KOIPE) mechanism.

As shown in the protocol flow, the PHO receives the final aggregate "count" (CNT), but the HC never sees the query parameters, and the PHO never sees individual health records. This "hidden in the crowd" principle ensures that an individual's status remains mathematically obscured unless a geographical grid contains only one person.

Experimental Evaluation

The scheme was evaluated using two real-life datasets from New York State: Lung Cancer Incidence (67,217 tumor incidences) and Birth Defect Data (over 1.2 million births).

Key results include:

  • Low Overhead: The communication cost for a user is approximately 1,294 bits, which is negligible for modern mobile devices.
  • Scalability: The PHO could process the Birth Defect dataset in about 1.72 hours for Poisson-based analysis on a single machine; cluster-based implementations would likely reduce this to minutes.
  • Statistical Accuracy: The privacy-preserving version of the Kulldorff scan produced the same statistical hotspots as the raw data version, proving that privacy does not compromise accuracy.

Conclusion

This framework demonstrates that ubiquitous cloud infrastructure can be leveraged for critical public health monitoring without sacrificing individual privacy. While the current model assumes "honest-but-curious" entities, future work may incorporate Differential Privacy to mitigate side-channel risks and explore incentive designs to encourage user participation. By decoupling identity from data and using advanced cryptographic primitives like KOIPE, the system provides a robust path toward secure, large-scale epidemiological tracking.

Summary updated

Abstract

The early detection of where and when fatal infectious diseases outbreak is of critical importance to the public health. To effectively detect, analyze and then intervene the spread of diseases, people's health status along with their location information should be timely collected. However, the conventional practices are via surveys or field health workers, which are highly costly and pose serious privacy threats to participants. In this paper, we for the first time propose to exploit the ubiquitous cloud services to collect users' multi-dimensional data in a secure and privacy-preserving manner and to enable the analysis of infectious disease. Specifically, we target at the spatial clustering analysis using Kulldorf scan statistic and propose a key-oblivious inner product encryption (KOIPE) mechanism to ensure that the untrusted entity only obtains the statistic instead of individual's data. Furthermore, we design an anonymous and sybil-resilient approach to protect the data collection process from double registration attacks and meanwhile preserve participant's privacy against untrusted cloud servers. A rigorous and comprehensive security analysis is given to validate our design, and we also conduct extensive simulations based on real-life datasets to demonstrate the performance of our scheme in terms of communication and computing overhead.
BibTeX
@inproceedings{liu2018cloud,
  author = {Jianqing Liu and Yaodan Hu and Hao Yue and Yanmin Gong and Yuguang Fang},
  title = {{A Cloud-Based Secure and Privacy-Preserving Clustering Analysis of Infectious Disease}},
  booktitle = {PAC},
  year = {2018},
  url = {https://openreview.net/forum?id=Unf_aQ97bUK},
}
Figure