Privacy-Preserving Explainable AIoT Application via SHAP Entropy Regularization

TL;DR

The paper tackles privacy leakage in post-hoc explainable AI for AIoT smart homes by introducing SHAP entropy regularization, which encourages more uniform SHAP attributions to obscure sensitive behavior signals. It formulates a SHAP-aware training objective and implements a SHAP entropy regularized LSTM for appliance-level energy forecasting. A suite of SHAP-based privacy attacks is developed to evaluate explanation-level leakage, and experiments on the REFIT dataset show substantial privacy gains with only minor utility loss compared to baseline and differential privacy baselines. The work advances trustworthy, privacy-preserving XAI for AIoT applications by balancing explanation fidelity with protection against inference attacks on SHAP outputs.

Abstract

The widespread integration of Artificial Intelligence of Things (AIoT) in smart home environments has amplified the demand for transparent and interpretable machine learning models. To foster user trust and comply with emerging regulatory frameworks, the Explainable AI (XAI) methods, particularly post-hoc techniques such as SHapley Additive exPlanations (SHAP), and Local Interpretable Model-Agnostic Explanations (LIME), are widely employed to elucidate model behavior. However, recent studies have shown that these explanation methods can inadvertently expose sensitive user attributes and behavioral patterns, thereby introducing new privacy risks. To address these concerns, we propose a novel privacy-preserving approach based on SHAP entropy regularization to mitigate privacy leakage in explainable AIoT applications. Our method incorporates an entropy-based regularization objective that penalizes low-entropy SHAP attribution distributions during training, promoting a more uniform spread of feature contributions. To evaluate the effectiveness of our approach, we developed a suite of SHAP-based privacy attacks that strategically leverage model explanation outputs to infer sensitive information. We validate our method through comparative evaluations using these attacks alongside utility metrics on benchmark smart home energy consumption datasets. Experimental results demonstrate that SHAP entropy regularization substantially reduces privacy leakage compared to baseline models, while maintaining high predictive accuracy and faithful explanation fidelity. This work contributes to the development of privacy-preserving explainable AI techniques for secure and trustworthy AIoT applications.

Figures (4)

• Figure 1: Comparing our SHAP-Regularized LSTM's Performance with Baseline LSTM and DP-LSTM Models.
• Figure 2: Comparing SHAP Attribution of Appliances (Features) with three different models.
• Figure 3: Comparing hourly SHAP entropy of appliances with three different models.
• Figure 4: Comparing entropy of each appliance across models: (a) aggregate SHAP entropy, (b) their difference with baseline LSTM.