Abstract
People-Centric Sensing (PCS) networks utilize mobile devices like smartphones to gather and share environmental or personal data. Unlike static sensor networks, PCS environments are highly dynamic and pose significant privacy risks. This paper presents the APP protocol (Attribute-based Private data sharing protocol), which employs Ciphertext-Policy Attribute-Based Encryption (CP-ABE) to provide fine-grained access control. By incorporating a novel Associative Data Index (ADI) and a two-tuple encryption strategy, the APP protocol ensures efficient data retrieval and privacy even when using "honest-but-curious" third-party storage.
Introduction
The rise of mobile computing has shifted the paradigm of sensing from static, embedded nodes to People-Centric Sensing (PCS). In PCS, individuals use their personal devices to contribute data to a collective pool. However, this mobility introduces two primary challenges: dynamic network topology and the privacy-utility trade-off. Traditional encryption methods struggle with the opportunistic nature of these connections and the need for scalable access control without managing thousands of individual public keys.
System Architecture
The proposed APP protocol operates within a multi-layered network model. It involves a Service Institution containing a Certificate Authority (CA) for key management and a Data Service Center (DSC) for storage. Mobile nodes act as Senders, Receivers, or opportunistic Relays that store and forward data when direct paths are unavailable.
The APP Protocol
To address the challenges of PCS, the APP protocol leverages CP-ABE. In this scheme, data is not encrypted for a specific identity but for an access policy defined by attributes (e.g., (Doctor AND Cardiology) OR (Researcher)). Only users with matching attributes can decrypt the information.
A key innovation is the "Two-Tuple Encryption" process. To minimize the computational and energy costs on mobile devices, the sender splits the data into two parts: a metadata ciphertext ($c_1$) and a payload ciphertext ($c_2$).
As detailed in Algorithm 2, the sender generates a random number $N$ and an attribute string $A_u$. These are bundled with the data to create $m_1$ and $m_2$, which are then encrypted into $c_1$ and $c_2$. This allows a receiver to first decrypt the smaller $c_1$ to verify relevance before committing to the more intensive decryption of the much larger $c_2$.
Furthermore, the Associative Data Index (ADI) allows the DSC to facilitate efficient data retrieval. Since the DSC cannot read the encrypted data, ADI provides a way for receivers to identify related data packets without a full linear scan, significantly reducing query overhead in large-scale datasets.
Performance Evaluation
The protocol was evaluated using an ARM-based experimental platform (S3C6410, 667MHz, ARM 11 Series) to test its suitability for mobile hardware:
- Latency: Key generation for 64 attributes took approximately 3.42 seconds. Decryption on the mobile hardware averaged 0.20 seconds per leaf node in the access policy tree.
- Overhead: For a policy with 32 leaf nodes, the transmission overhead was roughly 9.5 KB.
- Efficiency: In simulations with 12,000 data tuples where only 5% were relevant, the ADI scheme improved retrieval efficiency by nearly 50%, reducing the number of tuples a receiver needed to process from 12,000 down to 5,700.
Conclusion
The APP protocol provides a robust solution for private data sharing in dynamic PCS networks. By decoupling identity from access and optimizing retrieval through ADI and two-tuple encryption, it achieves a balance between security and performance on resource-constrained devices. Future research will focus on attribute revocation mechanisms and scaling the indexing scheme for massive-scale deployments.
Summary updated