Battery control defends smart meter data from unseen inference attacks
Learning Defensive Policies against Diverse Inference Attacks for Smart Meter Privacy
Machine Learning
Summary
Smart meters track how much electricity a home uses, but this detailed data can reveal what appliances are being used, which might invade privacy. The authors studied ways to hide these appliance signals by using batteries to change the overall energy use pattern without raising suspicion. They trained a system to create fake appliance usage patterns that confuse attackers trying to guess appliance use. Their method works well against many different guessing strategies, making smart meter data safer for users.
What this means in practice
- •For energy utility companies: Use battery-based policies to mask appliance use patterns in smart meter data and enhance customer privacy against unknown energy usage inference attacks.
- •For smart home device manufacturers: Incorporate learned load-shaping policies into home energy management systems to protect occupant privacy from appliance-level energy usage detection.
- •For home energy storage system vendors: Sell battery control software that hides appliance usage patterns to privacy-conscious consumers as a new value-added feature.$Commercial implications: Enables a privacy protection product layer for home battery systems that defends against appliance inference attacks on smart meter data.
Authors
Ruichang Zhang, Mustafa A. Mustafa
Abstract
Smart meter (SM) data provides fine-grained visibility into household energy consumption, but also exposes users to privacy risks. Inference attacks, known as non-intrusive load monitoring (NILM), can perform appliance-level inference from aggregate signals and recover sensitive behavioral patterns. In practice, attacker models are unknown and heterogeneous, making robust defense challenging. We formulate SM privacy protection as a black-box inference defense problem, aiming to reduce the recoverability of appliance-level information while generalizing across diverse and unseen attackers. We propose a proxy-guided hierarchical reinforcement learning framework that learns battery-based load-shaping policies to inject realistic but misleading appliance-level signatures into the aggregate signal, thereby disrupting the structured patterns exploited by NILM. A self-supervised aggregate-structure privacy probe provides a reconstruction-error-based surrogate reward for disrupting recoverable load structure, while a signature library makes the perturbations appliance-relevant and physically realizable through battery control. We provide theoretical rationale showing that proxy-guided optimization improves inference robustness under attacker diversity. Experiments on real-world datasets UK-DALE and REDD demonstrate strong cross-model and cross-appliance generalization. Across six unseen NILM attackers, covering four appliances on UK-DALE and five on REDD, our proposed defense increases average appliance-level RMSE by 107% and 166%, respectively, while reducing F1 score by 79% and 80%.