Abstract
This paper presents a privacy-preserving data aggregation scheme for smart grids, integrating Paillier homomorphic encryption with Shamir’s Secret Sharing. A novel "substitution strategy" is introduced to handle smart meter malfunctions, ensuring high data accuracy and system reliability without compromising user privacy. Simulations demonstrate that the proposed scheme maintains an error rate as low as 0.02%, significantly outperforming existing protocols like DG-APED in fault-tolerant scenarios.
Introduction
The modernization of power grids into "smart grids" relies heavily on high-frequency, real-time data from smart meters (SMs) to manage electricity loads and implement dynamic pricing. However, this granularity exposes sensitive user habits, such as daily routines and occupancy patterns. While data aggregation—summing usage across groups—is a common privacy-preserving technique, it remains vulnerable to differential attacks and often lacks robustness against device malfunctions. If a single meter in an aggregation group fails, many traditional schemes cannot correctly decrypt the total sum, leading to data loss or significant errors.
Proposed System Model
The authors propose a multi-layered cryptographic scheme (referred to as OS) designed to balance privacy with operational reliability. The architecture involves several key entities: Smart Meters (SMs), a Key Initialization Center (KIC), a Data Aggregation device (DA), and a Control Center (CC).
The core components of the OS scheme include:
- Paillier Cryptosystem: An additive homomorphic encryption method that allows the DA to sum encrypted meter readings without decrypting them.
- Shamir’s Secret Sharing: The decryption key is distributed as shares among users. The CC can only reconstruct the key if a threshold $d$ of shares is met, preventing "sum-minus-sum" differential attacks.
- Substitution Strategy for Error Tolerance: When a meter fails, the DA identifies it using hash functions and substitutes its missing value with data from a "twin" meter in a different group. This twin meter is initialized with identical secret parameters, leveraging the log-normal distribution of residential power consumption to maintain statistical accuracy.
Evaluation and Results
The performance of the OS scheme was evaluated through numerical simulations involving 4,000 to 10,000 smart meters and compared against the DG-APED protocol.
The primary advantage of the OS scheme lies in its resilience to meter malfunctions. As shown in Figure 4, while the error rate of DG-APED spikes dramatically as the ratio of malfunctioning meters ($M/N$) increases, the OS scheme maintains a remarkably low and stable error rate.
Key findings include:
- High Accuracy: The OS scheme achieved an average error rate of 0.02%, peaking at only 0.07% in extreme failure scenarios.
- Scalability: Computational complexity for the OS scheme scales linearly and remains lower than that of DG-APED, making it more suitable for large-scale urban deployments.
- Robustness: The system remains operational until the malfunction rate reaches approximately 97%, at which point a key reassignment is required.
Conclusion
The proposed OS scheme effectively addresses the dual challenges of privacy and reliability in smart grid data aggregation. By combining homomorphic encryption with a novel secret-sharing-based substitution strategy, the system ensures that utility companies can access accurate aggregate data even in the presence of faulty hardware, all while keeping individual user data strictly confidential. Future work will explore integrating this real-time privacy framework with individual billing systems.
Summary updated