Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Make Your Home Safe: Time-aware Unsupervised User Behavior Anomaly Detection in Smart Homes via Loss-guided Mask

Jingyu Xiao, Zhiyao Xu, Qingsong Zou, Qing Li, Dan Zhao, Dong Fang, Ruoyu Li, Wenxin Tang, Kang Li, Xudong Zuo, Penghui Hu, Yong Jiang, Zixuan Weng, Michael R. Lyv

TL;DR

SmartGuard tackles unsupervised anomaly detection in smart-home user behaviors by addressing three core challenges: imbalanced behavior frequencies, lack of temporal context, and noise interference. It introduces a triad of innovations—Loss-guided Dynamic Mask Strategy (LDMS), Three-level Time-aware Position Embedding (TTPE), and Noise-aware Weighted Reconstruction Loss (NWRL)—implemented on an autoencoder-Transformer backbone. The approach yields a weighted reconstruction score for sequences, achieving superior performance over diverse baselines across three real-world datasets and ten anomaly types, while also providing interpretable attention and embedding insights. This work has practical implications for real-time smart-home security, offering robust detection with explainable behavior representations.

Abstract

Smart homes, powered by the Internet of Things, offer great convenience but also pose security concerns due to abnormal behaviors, such as improper operations of users and potential attacks from malicious attackers. Several behavior modeling methods have been proposed to identify abnormal behaviors and mitigate potential risks. However, their performance often falls short because they do not effectively learn less frequent behaviors, consider temporal context, or account for the impact of noise in human behaviors. In this paper, we propose SmartGuard, an autoencoder-based unsupervised user behavior anomaly detection framework. First, we design a Loss-guided Dynamic Mask Strategy (LDMS) to encourage the model to learn less frequent behaviors, which are often overlooked during learning. Second, we propose a Three-level Time-aware Position Embedding (TTPE) to incorporate temporal information into positional embedding to detect temporal context anomaly. Third, we propose a Noise-aware Weighted Reconstruction Loss (NWRL) that assigns different weights for routine behaviors and noise behaviors to mitigate the interference of noise behaviors during inference. Comprehensive experiments on three datasets with ten types of anomaly behaviors demonstrates that SmartGuard consistently outperforms state-of-the-art baselines and also offers highly interpretable results.

Make Your Home Safe: Time-aware Unsupervised User Behavior Anomaly Detection in Smart Homes via Loss-guided Mask

TL;DR

SmartGuard tackles unsupervised anomaly detection in smart-home user behaviors by addressing three core challenges: imbalanced behavior frequencies, lack of temporal context, and noise interference. It introduces a triad of innovations—Loss-guided Dynamic Mask Strategy (LDMS), Three-level Time-aware Position Embedding (TTPE), and Noise-aware Weighted Reconstruction Loss (NWRL)—implemented on an autoencoder-Transformer backbone. The approach yields a weighted reconstruction score for sequences, achieving superior performance over diverse baselines across three real-world datasets and ten anomaly types, while also providing interpretable attention and embedding insights. This work has practical implications for real-time smart-home security, offering robust detection with explainable behavior representations.

Abstract

Smart homes, powered by the Internet of Things, offer great convenience but also pose security concerns due to abnormal behaviors, such as improper operations of users and potential attacks from malicious attackers. Several behavior modeling methods have been proposed to identify abnormal behaviors and mitigate potential risks. However, their performance often falls short because they do not effectively learn less frequent behaviors, consider temporal context, or account for the impact of noise in human behaviors. In this paper, we propose SmartGuard, an autoencoder-based unsupervised user behavior anomaly detection framework. First, we design a Loss-guided Dynamic Mask Strategy (LDMS) to encourage the model to learn less frequent behaviors, which are often overlooked during learning. Second, we propose a Three-level Time-aware Position Embedding (TTPE) to incorporate temporal information into positional embedding to detect temporal context anomaly. Third, we propose a Noise-aware Weighted Reconstruction Loss (NWRL) that assigns different weights for routine behaviors and noise behaviors to mitigate the interference of noise behaviors during inference. Comprehensive experiments on three datasets with ten types of anomaly behaviors demonstrates that SmartGuard consistently outperforms state-of-the-art baselines and also offers highly interpretable results.
Paper Structure (36 sections, 20 equations, 15 figures, 8 tables)

This paper contains 36 sections, 20 equations, 15 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. User Behavior Modeling in Smart Homes
  4. Attacks and Defenses in Smart Homes
  5. Problem Formulation
  6. Methodology
  7. Solution Overview
  8. Loss-guided Dynamic Mask Strategy
  9. Autoencoder with Temporal Information
  10. Three-level Time-aware Positional Encoder
  11. Sequence Encoder
  12. Sequence Decoder
  13. Objective Function
  14. Noise-aware Weighted Reconstruction Loss
  15. Experiments
  16. ...and 21 more sections

Figures (15)

  • Figure 1: Reconstruction losses for behaviors with different occurrence frequencies.
  • Figure 2: Example of three user behaviors with the same behavior order. Sequence 2 and 3 are abnormal due to their inappropriate timing and excessive duration.
  • Figure 3: An example of noise behaviors.
  • Figure 4: Different types of anomaly behaviors.
  • Figure 5: The overview of SmartGuard.
  • ...and 10 more figures

Theorems & Definitions (2)

  • Definition 1
  • Definition 2