Outsourceable Privacy-Preserving Power Usage Control in a Smart Grid

Hu Chun, Kui Ren, Wei JiangDBSec 2015 · DOI

TL;DR

This paper introduces a novel protocol for privacy-preserving power usage control (P-PUC) in smart grids that allows computations to be outsourced to cloud servers. This significantly reduces the computational burden on individual households and utility companies, while maintaining strong privacy guarantees.

What Problem Does This Paper Address?

Smart grids offer numerous benefits, including improved efficiency and reliability, by enabling utility companies to monitor power consumption. However, collecting granular power usage data from smart meters raises significant privacy concerns. This data can reveal intimate details about individuals' daily lives, making them vulnerable to misuse. To address this, privacy-preserving power usage control (P-PUC) protocols have been developed. The challenge is that existing P-PUC protocols are often computationally intensive for households and utility companies, and they don't effectively leverage the cost-efficiency and operational flexibility of cloud computing. This paper aims to bridge this gap by proposing an outsourceable P-PUC protocol that offloads computation to the cloud, making it more practical and efficient.

Methods

The proposed Outsourceable Privacy-Preserving Power Usage Control (OP-PUC) protocol utilizes a multi-server framework with two independent cloud servers (S1 and S2) to handle computations. The core idea is to hide sensitive inputs (individual power consumption a_i and the utility's threshold t) from the cloud servers. This is achieved through additive secret sharing, where each input is split into shares distributed between S1 and S2.

The protocol operates in three stages:

  1. Data Collection: Each household P_i and the utility company U use a Secure_Split protocol to generate secret shares of their respective inputs (a_i and t). These shares are then sent to S1 and S2. The cloud servers then compute the aggregated secret shares of the total neighborhood power consumption (a' = Σa'_i and a'' = Σa''_i).
  2. Secure Comparison: To determine if the total power consumption a exceeds the threshold t, a garbled circuit is employed for secure comparison. This allows S1 and S2 to compare their secret shares of a and t without revealing the actual values.
  3. Power Reduction Calculation: If a > t, the protocol proceeds to calculate the required power reduction δ_i for each household. Two strategies are proposed:
  4. Strategy 1 (Secure Maximum): Identifies the user with the highest consumption and orders them to reduce power. This is also implemented using a garbled circuit.
  5. Strategy 2 (Secure Division): Calculates a specific power reduction amount for each user based on their consumption relative to the threshold. This involves a secure division operation, also implemented using a garbled circuit based on a "shift and subtract" non-restoring method.

The use of garbled circuits is a key technique, enabling efficient and secure computation over encrypted data, which is crucial for outsourcing to the cloud.

Datasets

Not discussed in the paper. The paper focuses on a theoretical protocol and its performance evaluation through simulation.

Key Insights and Ideas

  • Cloud-Native P-PUC: The most significant insight is the feasibility and efficiency of outsourcing P-PUC computations to the cloud. This paradigm shift can dramatically reduce the burden on end-users and utility providers.
  • Additive Secret Sharing for Input Hiding: Using additive secret sharing is an efficient way to hide input values from cloud servers, which is a critical requirement for privacy.
  • Garbled Circuits for Secure Computation: The paper effectively leverages garbled circuits for complex operations like secure comparison and division, which are essential for P-PUC but difficult to implement efficiently in an outsourced setting with encrypted data.
  • Scalability and Efficiency: The proposed protocol demonstrates good scalability with the number of households and is more efficient than existing P-PUC protocols, especially for strategy 2.

Key Results

  • Computational Efficiency: The OP-PUC protocol incurs negligible computational costs for individual users and the utility company, as almost all computations are outsourced to the cloud servers.
  • Scalability: The protocol shows good scalability with the number of households. For strategy 2 (secure division), the running time of cloud servers remains nearly constant as the number of households increases.
  • Performance Comparison:
    • OP-PUC (strategy 1) shows performance close to existing methods (STPUCmax) but is more scalable with domain size.
  • OP-PUC² (strategy 2) is significantly more efficient and scalable than existing secure division protocols (STPUCdiv), offering a speedup of 3 to 4 times.
  • Communication Complexity: The total communication complexity is bounded by O(n log N) bits, which is considered very small.

Limitations and Future Work

  • Semi-Honest Model: The security proof is based on the semi-honest model, where participants follow the protocol but may try to learn information from it. The authors acknowledge that a malicious model (where participants can deviate arbitrarily) would require further investigation.
  • Two Cloud Servers: The current protocol uses two cloud servers. Future work could explore using three or more cloud servers to potentially improve computation costs further by enabling computations on secure shares directly.
  • Malicious Model Security: The authors explicitly state their intention to develop OP-PUC protocols secure under the malicious model, possibly by adopting threshold homomorphic encryption or Shamir secret sharing.
  • Further Optimization: Investigating the pros and cons of different secret sharing schemes for improving efficiency is also a future direction.

Summary updated

BibTeX
@inproceedings{chun2015outsourceable,
  author = {Hu Chun and Kui Ren and Wei Jiang},
  title = {{Outsourceable Privacy-Preserving Power Usage Control in a Smart Grid}},
  booktitle = {DBSec},
  year = {2015},
  doi = {10.1007/978-3-319-20810-7\_8},
}
Figure