TL;DR
This paper introduces PMHA-DP, a novel scheme for privacy-preserving health data aggregation in cloud-assisted wireless body area networks (WBANs). It addresses limitations in existing methods by offering enhanced privacy guarantees through differential privacy, supporting a wider range of aggregation functions (including weighted average and histogram), and ensuring fault tolerance.
What Problem Does This Paper Address?
The proliferation of health data from WBANs presents a challenge for storage and computation, especially with resource-constrained devices. Outsourcing this data to the cloud is appealing, but raises concerns about data security and user privacy. Existing cloud-assisted WBAN aggregation schemes often struggle with:
- Limited Functionality: Many schemes only support basic aggregation like summation or additive aggregation, failing to address more complex needs like weighted averages or histograms.
- Privacy Leaks: Some methods allow cloud servers to learn sensitive aggregated data (e.g., the sum of all health data), posing a risk if servers are compromised.
- Lack of Robustness: Fault tolerance, crucial in dynamic WBAN environments, is not always adequately addressed.
- Differential Privacy Gaps: While some schemes offer privacy, they may not fully meet the stringent mathematical guarantees of differential privacy, especially for complex aggregations.
This paper aims to bridge these gaps by proposing a comprehensive solution that is multifunctional, privacy-enhanced, and fault-tolerant.
Methods
The proposed PMHA-DP scheme leverages several key cryptographic and privacy-preserving techniques:
- Boneh-Goh-Nissim (BGN) Cryptosystem: This homomorphic encryption scheme is used to encrypt health data. Its additive and multiplicative properties are crucial for performing computations on encrypted data without decryption.
- Differential Privacy (DP): Specifically, the Laplace mechanism is employed to add calibrated noise to the final aggregation results. This ensures that the output is differentially private, meaning an adversary cannot infer information about any single individual's data by observing the aggregated result.
- Multifunctional Aggregation:
- Basic Average Aggregation Scheme (BAAS): Computes the average of health data.
- Weighted Average Aggregation Scheme (WAAS): Introduces weighted averaging, allowing for more objective aggregation by considering data reliability. This is a novel contribution in this context.
- Privacy-Enhanced Average Aggregation Scheme (PAAS): Addresses the issue of cloud servers learning the sum of data. It uses additional private keys to encrypt the data twice, making the aggregated sum unreadable to the cloud servers.
- Hierarchical Method for Histogram (HMH): Implements histogram aggregation, a common statistical tool, and uses post-processing techniques to improve accuracy while maintaining differential privacy.
- Fault Tolerance: The system is designed to withstand the compromise of up to
l = [n/2] - 1cloud servers, ensuring continuous operation even with partial server failures.
Datasets
Not discussed in the paper. The paper focuses on theoretical contributions and simulations rather than experiments on real-world datasets.
Key Insights and Ideas
- Weighted Average is Crucial: The paper highlights that weighted averaging is a practical and often preferred aggregation method in real-world scenarios, and its absence in prior work was a significant gap.
- Separation of Concerns for Privacy: By assigning the final aggregation and noise addition to a trusted authority (TA) rather than cloud servers, PMHA-DP significantly enhances privacy. This prevents cloud servers from learning even aggregated sensitive information.
- Dual Encryption for Enhanced Privacy: The PAAS scheme's use of dual encryption with additional private keys is a clever way to obscure the aggregated sum from potentially curious cloud servers.
- Post-Processing for Accuracy: The application of post-processing techniques (like weighted averaging and mean consistency) to histogram aggregation is key to balancing differential privacy with the need for accurate results.
Key Results
- Enhanced Privacy: PMHA-DP provides strong privacy guarantees, resisting various adversary models, including compromised communication channels, compromised cloud servers, and differential attacks on the trusted authority.
- Multifunctionality: The scheme supports additive (average, weighted average) and non-additive (histogram) aggregations, making it versatile for various healthcare applications.
- Fault Tolerance: The system can tolerate failures of a significant portion of cloud servers and individual user data submission failures.
- Reduced Communication Overhead: Performance evaluations show that PMHA-DP's non-additive aggregation scheme (NAS) and PAAS lead to less communication overhead compared to existing schemes like MHDA+.
- Acceptable Computational Overhead: While PAAS incurs higher computational cost due to dual encryption, it is deemed acceptable for the enhanced privacy it provides, especially in big data environments where parallel processing can mitigate the impact.
Limitations and Future Work
- Data Tampering: The paper acknowledges that detecting and handling tampered data sent by compromised users is a significant challenge that remains an open problem.
- Efficiency for Large Data: The Boneh-Goh-Nissim cryptosystem, while suitable for short messages, might not be efficient for encrypting very large files (e.g., video data).
- Future Work: The authors plan to explore data aggregation techniques in big data environments, investigate other aggregation functions, and improve the scheme's efficiency. They also aim to address the data tampering issue.
Summary updated