Adaptive Anomaly Detection for Internet of Things in Hierarchical Edge Computing: A Contextual-Bandit Approach

TL;DR

This paper presents an adaptive anomaly detection method for IoT in hierarchical edge computing, using a contextual-bandit reinforcement learning approach to balance accuracy and delay effectively.

Contribution

It introduces a novel adaptive model selection scheme formulated as a contextual-bandit problem, improving anomaly detection in IoT with hierarchical edge computing.

Findings

Achieves the best accuracy-delay tradeoff on univariate datasets.

Attains highest accuracy and F1-score on multivariate datasets.

Demonstrates effectiveness on real IoT device testbed.

Abstract

The advances in deep neural networks (DNN) have significantly enhanced real-time detection of anomalous data in IoT applications. However, the complexity-accuracy-delay dilemma persists: complex DNN models offer higher accuracy, but typical IoT devices can barely afford the computation load, and the remedy of offloading the load to the cloud incurs long delay. In this paper, we address this challenge by proposing an adaptive anomaly detection scheme with hierarchical edge computing (HEC). Specifically, we first construct multiple anomaly detection DNN models with increasing complexity, and associate each of them to a corresponding HEC layer. Then, we design an adaptive model selection scheme that is formulated as a contextual-bandit problem and solved by using a reinforcement learning policy network. We also incorporate a parallelism policy training method to accelerate the training process by taking advantage of distributed models. We build an HEC testbed using real IoT devices, implement and evaluate our contextual-bandit approach with both univariate and multivariate IoT datasets. In comparison with both baseline and state-of-the-art schemes, our adaptive approach strikes the best accuracy-delay tradeoff on the univariate dataset, and achieves the best accuracy and F1-score on the multivariate dataset with only negligibly longer delay than the best (but inflexible) scheme.