Abstract
As cloud computing facilitates widespread data sharing for research in artificial intelligence and data analysis, protecting sensitive information remains a critical challenge. This paper proposes an identity-based auditing scheme for shared cloud data that incorporates a secure mechanism for hiding sensitive information. The scheme allows users to share plaintext data with researchers while keeping sensitive identifiers invisible to both the cloud and the researchers. By utilizing a novel blinding mechanism and a simplified signature algorithm, the proposed system prevents malicious managers from tampering with data and significantly improves computational efficiency compared to existing state-of-the-art approaches.
The Challenge of Privacy in Shared Cloud Data
The proliferation of cloud-based research platforms has made sharing large datasets, such as Electronic Health Records (EHRs), more efficient than ever. However, these datasets often contain sensitive personal information that must be protected to ensure patient privacy.
In traditional models, a "manager" (such as a hospital administrator) is often responsible for sanitizing data before it is uploaded. This creates a "malicious manager" vulnerability: if the manager has the power to modify signatures to hide information, they also possess the capability to forge data or conceal errors. Additionally, many existing schemes require users to maintain local records of random "blinding" values to recover their original data, leading to potential data loss if local storage fails.
Proposed Identity-Based Auditing Scheme
To address these limitations, the authors developed an Identity-Based Auditing (IBA) scheme designed for high concurrency and massive data scenarios. The architecture introduces several key innovations to balance privacy, integrity, and usability.
Key Technical Components
- Identity-Based Cryptography (IBC): The system uses a user's identity (e.g., an email address) as their public key. This eliminates the overhead of managing complex digital certificates required in traditional Public Key Infrastructure (PKI).
- Deterministic Blinding Mechanism: Rather than relying on random numbers, the scheme derives blinding factors from the user’s unique private key ($sk_{ID}$). This ensures that the data owner can always recover the original plaintext without needing to store any additional local metadata.
- Hess’s Signature Scheme: The authors adapted this efficient identity-based signature algorithm to sign data blocks, ensuring that only the data owner can generate valid signatures while allowing for efficient verification.
- Malicious Manager Prevention: The system model is specifically designed to decouple the ability to inspect data from the ability to sign it. While a manager can verify that sensitive fields are blinded, they do not have the cryptographic keys necessary to modify the data or its signatures. Any tampering by a manager is detectable during the auditing phase.
Performance and Experimental Results
The proposed scheme was evaluated using a dataset of 200 sample EHRs from the UESTC Hospital. The experimental results highlight significant performance gains over previous methods:
- Computational Efficiency: Signature generation and verification processes were found to be approximately 50% more efficient. For a standard 2KB file, the entire pipeline—including blinding, signing, and uploading—takes less than 2 seconds.
- High-Speed Hiding: The system processes sensitive information at a rate of roughly 72KB/s. In tests, hiding 60KB of sensitive data across 200 records was completed in just 830.3ms.
- Scalable Auditing: The Third Party Auditor (TPA) can challenge and verify 1,000 data blocks in under 1.5 seconds, demonstrating the system's readiness for large-scale cloud deployments.
- Zero Storage Overhead: By using the private key as a seed for blinding, the system removes the requirement for users to store extra metadata locally, enhancing both security and convenience.
Conclusion
This research provides a robust solution for the secure sharing of sensitive data in the cloud. By integrating identity-based signatures with a novel private-key-driven blinding mechanism, the authors have created a system that is both more efficient and more secure than previous models. The prevention of the "malicious manager" problem is a particularly significant contribution, ensuring data integrity in hierarchical organizational structures. Future research will explore extending these auditing capabilities to fully dynamic data environments.
Summary updated