TL;DR
This paper introduces PFRB, a novel blockchain system for IoT that allows for fine-grained data redaction while crucially hiding users' redaction policies. By combining Newton's interpolation formula and multi-authority attribute-based encryption, PFRB addresses critical privacy concerns in sensitive IoT applications like healthcare, enabling secure data management without exposing private policy details.
What Problem Does This Paper Address?
The Internet of Things (IoT) generates vast amounts of data, necessitating secure and manageable storage. While blockchains offer immutability and traceability, this rigidity conflicts with regulations like GDPR's "right to be forgotten." Redactable blockchains aim to solve this by allowing data modification. However, existing fine-grained redactable blockchain solutions suffer from a significant privacy flaw: they expose users' redaction policies. These policies often contain highly sensitive personal information (e.g., health conditions, geographical locations), making them unsuitable for privacy-sensitive IoT applications. Leaking this policy information can lead to discrimination and attacks. This paper addresses the urgent need for a decentralized, fine-grained redactable blockchain that also protects the privacy of users' redaction policies.
Methods
The proposed system, PFRB (Privacy-Preserving Redactable Blockchain), builds upon existing fine-grained redactable blockchain techniques but introduces a crucial policy-hiding mechanism. The core innovation lies in how PFRB handles redaction policies:
- Policy Hiding via Secret Sharing: PFRB leverages Newton's interpolation formula-based secret sharing. User policies are converted into polynomial parameters. This allows for policy matching and redaction without directly revealing the policy content. Only authorized entities, after successful policy matching, can reconstruct the necessary information.
- Attribute-Based Encryption (ABE): To further obscure these polynomial parameters, PFRB employs multi-authority attribute-based encryption. This ensures that even if the parameters are intercepted, their meaning and utility are hidden unless the correct attributes are possessed.
- Decentralized Chameleon Hashing: PFRB utilizes decentralized chameleon hashes, a standard technique for redactable blockchains, to enable the modification of data entries.
The system involves four types of entities: Authorities (trusted, initialize system, issue attributes), Transaction Owners (trusted, submit data), Transaction Modifiers (users who can redact data based on policies), and Blockchain Participants (verify transactions).
Datasets
The experiments were conducted using the MNH9 dataset, which consists of millions of news headlines. This dataset was provided by the Australian Broadcasting Corporation and is a real-world open dataset.
Key Insights and Ideas
- Policy Privacy is Paramount in IoT: The paper strongly emphasizes that for many IoT applications (especially healthcare and smart industry), the privacy of the rules for data modification (the policies) is as critical, if not more so, than the privacy of the data itself.
- Newton's Interpolation as a Privacy Tool: The clever application of Newton's interpolation formula for secret sharing transforms sensitive policy information into a form that can be manipulated and matched without direct exposure. This is a novel use of the technique for policy obfuscation.
- Layered Security for Policy Hiding: Combining secret sharing with multi-authority ABE creates a robust defense against policy leakage, ensuring that only authorized parties can infer policy details and perform redactions.
- Decentralization is Key for IoT: The design adheres to decentralized principles, which are inherent to many IoT deployments and crucial for scalability and resilience.
Key Results
The experimental evaluation on the FISCO blockchain platform demonstrated that PFRB achieves competitive efficiency compared to existing redactable blockchain schemes.
- Key Generation: PFRB shows significant improvement, with required time increasing linearly with the number of attributes. It outperforms traditional approaches and RBDS22.
- Hashing: PFRB exhibits a notable efficiency gain, with required time increasing linearly with policy size, unlike traditional approaches that show exponential growth.
- Adaptation: While all schemes show exponential growth with policy size, PFRB performs better than traditional methods and RBDS22, especially for smaller policy sets.
- Verification: The response time for verification is linear with the number of requests and shows no significant difference across the compared schemes, indicating comparable performance in this aspect.
Limitations and Future Work
- Limitations: The paper acknowledges that the adaptation algorithm's performance can be challenged by the enumeration of attribute-value pairs. While PFRB outperforms traditional methods, its performance can still be exponential with the number of policies in this specific operation.
- Future Work: The authors plan to focus on providing more comprehensive privacy protection (e.g., data privacy) and richer functionalities like accountability and revocation in decentralized IoT systems. Additionally, they aim to enhance data analytic mechanisms by integrating machine learning schemes.
Summary updated