Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Online Continual Learning for Anomaly Detection in IoT under Data Distribution Shifts

Matea Marinova, Shashi Raj Pandey, Junya Shiraishi, Martin Voigt Vejling, Valentin Rakovic, Petar Popovski

TL;DR

OCLADS keeps track of data distribution shifts to timely update the on-device IoT AD model and introduces two mechanisms during the interaction between the resource-constrained IoT device and an edge server (ES).

Abstract

In this work, we present OCLADS, a novel communication framework with continual learning (CL) for Internet of Things (IoT) anomaly detection (AD) when operating in non-stationary environments. As the statistical properties of the observed data change with time, the on-device inference model becomes obsolete, which necessitates strategic model updating. OCLADS keeps track of data distribution shifts to timely update the on-device IoT AD model. To do so, OCLADS introduces two mechanisms during the interaction between the resource-constrained IoT device and an edge server (ES): i) an intelligent sample selection mechanism at the device for data transmission, and ii) a distribution-shift detection mechanism at the ES for model updating. Experimental results with TinyML demonstrate that our proposed framework achieves high inference accuracy while realizing a significantly smaller number of model updates compared to the baseline schemes.

Online Continual Learning for Anomaly Detection in IoT under Data Distribution Shifts

TL;DR

OCLADS keeps track of data distribution shifts to timely update the on-device IoT AD model and introduces two mechanisms during the interaction between the resource-constrained IoT device and an edge server (ES).

Abstract

In this work, we present OCLADS, a novel communication framework with continual learning (CL) for Internet of Things (IoT) anomaly detection (AD) when operating in non-stationary environments. As the statistical properties of the observed data change with time, the on-device inference model becomes obsolete, which necessitates strategic model updating. OCLADS keeps track of data distribution shifts to timely update the on-device IoT AD model. To do so, OCLADS introduces two mechanisms during the interaction between the resource-constrained IoT device and an edge server (ES): i) an intelligent sample selection mechanism at the device for data transmission, and ii) a distribution-shift detection mechanism at the ES for model updating. Experimental results with TinyML demonstrate that our proposed framework achieves high inference accuracy while realizing a significantly smaller number of model updates compared to the baseline schemes.
Paper Structure (14 sections, 6 equations, 2 figures)

This paper contains 14 sections, 6 equations, 2 figures.

Table of Contents

  1. Introduction
  2. System Model
  3. Observations Model with Distribution Shift
  4. Anomaly Process and its Detection
  5. Distribution Shifts
  6. Communication Model
  7. Proposed Framework
  8. Sample Selection and Uplink Data Transmission
  9. Hypothesis Testing Framework for Model Updates
  10. Experimental Setup
  11. Datasets and Models
  12. Simulation Model and Baseline Schemes
  13. Numerical Results
  14. Conclusion

Figures (2)

  • Figure 1: Overview of the OCLADS framework for anomaly detection in online under non-stationary conditions.
  • Figure 2: Average online macro F1-scores w.r.t. the number of communication rounds and device-side model updates.