A Privacy-Preserving Framework Using Homomorphic Encryption for Smart Metering Systems

Weiyan Xu, Jack Sun, Rachel Cardell-Oliver, Ajmal Mian, Jin B. Hong2023 · DOI

TL;DR

This paper introduces a novel privacy-preserving framework for smart metering systems (SMSs) that leverages homomorphic encryption (HE) to enable computations on encrypted data. By defining trust boundaries, the framework offers flexible privacy protection across different SMS scenarios, demonstrating its feasibility with real-world data and acceptable computational overhead.

What Problem Does This Paper Address?

Smart metering systems (SMSs) generate vast amounts of granular data about energy and resource consumption. While this data is valuable for utilities and consumers (e.g., for billing, load forecasting, and efficiency improvements), it also poses significant privacy risks. This data can be used to infer sensitive information about individuals' daily routines, habits, and even presence, making them vulnerable to surveillance and misuse by malicious actors, including service providers, third-party analysts, or even neighbors. Existing privacy-preserving techniques often fall short, either by reducing data utility (e.g., differential privacy) or having limitations in practical deployment. The paper aims to address this by proposing a robust and adaptable privacy framework for SMSs.

Methods

The core of the proposed framework is homomorphic encryption (HE), specifically the CKKS scheme, which allows computations to be performed directly on encrypted data without decryption. This is crucial for privacy as it means sensitive consumption data never needs to be exposed in its plaintext form to third parties.

A key innovation is the introduction of trust boundaries (TBs). TBs are logical or physical boundaries that define domains with different levels of trust. By analyzing SMS scenarios through the lens of TBs, the authors design configurations for HE, including key management and distribution, tailored to specific trust relationships between entities (clients, service providers, and analysts).

The framework is designed to be flexible and can be deployed in three scenarios, each with different TB configurations:

  1. Scenario 1: Households and Service Provider within the trust boundary.
  2. Scenario 2: Households within the trust boundary, but the Service Provider is outside.
  3. Scenario 3: Only the household is within its own trust boundary.

For practical implementation, the authors utilize the Lattigo HE library, chosen for its performance and support for multiparty computation (MPC) protocols, which are essential for Scenario 3. The framework implements two common computation metrics: summation (for billing and monitoring) and variance (for load forecasting and demand management), demonstrating HE's ability to handle these analytical tasks on encrypted data.

Datasets

The paper uses a real-world Smart Water Metering (SWM) dataset from the DAIAD project. This dataset contains time-series data for 1007 consumers, with approximately 16.8 million measurements. For evaluation, the data was grouped into 8640 water consumption records per user per day.

Key Insights and Ideas

  • Trust Boundaries as a Design Principle: Defining trust boundaries is a powerful way to systematically design privacy-preserving systems for complex scenarios like SMS. It allows for tailored HE configurations rather than a one-size-fits-all approach.
  • HE for Practical SMS Analytics: The paper demonstrates that HE, while computationally intensive, can be practically applied to perform essential analytical tasks like summation and variance on encrypted smart meter data, offering a strong privacy guarantee.
  • Flexibility through Scenarios: The three proposed scenarios, based on varying trust assumptions, highlight the adaptability of the HE framework to different organizational and security contexts.
  • Key Distribution is Critical: The security of HE heavily relies on secure key management. The paper emphasizes the importance of using Certificate Authorities (CAs) and secure protocols like TLS to ensure the integrity and authenticity of keys exchanged between entities.

Key Results

The paper evaluates the proposed framework's performance in terms of time and disk consumption. For a sample size of 100 households, the HE computations for summation and variance took 58,235 ms and 127,423 ms, respectively. The authors deem this computational overhead "acceptable from a cost-benefit perspective while ensuring data privacy."

Key findings from the evaluation include:

  • Rotation key generation is the most time-consuming operation and requires the largest disk storage.
  • HE multiplication is a significant contributor to the overall computation time for analysts.
  • Memory consumption for the service provider and households generally increases linearly with the number of households, while analyst memory remains relatively stable.
  • The amortized time for encryption and decryption operations in their framework (using Lattigo) is comparable to other HE libraries like SEAL.

Limitations and Future Work

  • Computational Overhead: While deemed acceptable, HE still incurs significant computational overhead compared to plaintext computations. The paper acknowledges that this can be a barrier for some applications.
  • Complexity of Advanced Metrics: The current implementation focuses on simple metrics (summation, variance). The paper notes that converting more complex analytical metrics (e.g., load forecasting) into HE computations can be challenging due to limitations in current HE library primitives. Future work could explore how to integrate HE for parts of these complex metrics.
  • Integrity Preservation: The framework primarily focuses on privacy. The authors suggest that integrating Distributed Ledger Technology (DLT) could further enhance data integrity, preventing tampering, and propose this as a direction for future research.
  • Energy Consumption: The paper acknowledges that HE deployment can have significant energy consumption implications, which warrants further investigation and optimization, potentially through specialized hardware.
  • Adversary Assumptions: While the framework addresses various adversary types, the authors note that for complete data integrity guarantees, especially against malicious type 2 adversaries (service providers), additional mechanisms beyond HE might be needed.

Summary updated

BibTeX
@misc{xu2023privacy,
  author = {Weiyan Xu and Jack Sun and Rachel Cardell-Oliver and Ajmal Mian and Jin B. Hong},
  title = {{A Privacy-Preserving Framework Using Homomorphic Encryption for Smart Metering Systems}},
  year = {2023},
  doi = {10.3390/s23104746},
}
Figure