Learning Graph Structures With Transformer for Multivariate Time-Series Anomaly Detection in IoT

Обучение структур графов с помощью Transformer для обнаружения аномалий в многомерных временных рядах в Интернете вещей
Zixuan Yuan, Xiuzhen Cheng, Dingshuo Chen, Xiao Zhang, Zekai Chen, Xiao Zhang
2021-07-27

Gumbel-softmax graph learningInfluence Propagation convolutionIoT sensor networksTransformer-based architecturemultivariate time series anomaly detection
Many real-world Internet of Things (IoT) systems, which include a variety of Internet-connected sensory devices, produce substantial amounts of multivariate time-series data. Meanwhile, vital IoT infrastructures, such as smart power grids and water distribution networks are frequently targeted by cyberattacks, making anomaly detection an important study topic. Modeling such relatedness is, nevertheless, unavoidable for any efficient and effective anomaly detection system, given the intricate topological and nonlinear connections that are originally unknown among sensors. Furthermore, detecting anomalies in multivariate time series is difficult due to their temporal dependency and stochasticity. This article presented GTA, a new framework for multivariate time-series anomaly detection that involves automatically learning a graph structure, graph convolution, and modeling temporal dependency using a transformer-based architecture. The connection learning policy, which is based on the Gumbel-softmax sampling approach to learn bidirected links among sensors directly, is at the heart of the learning graph structure. To describe the anomaly information flow between network nodes, we introduced a new graph convolution called influence propagation convolution. In addition, to tackle the quadratic complexity barrier, we suggested a multibranch attention mechanism to replace the original multihead self-attention method. Extensive experiments on four publicly available anomaly detection benchmarks further demonstrate the superiority of our approach over alternative state of the arts.
1
A multi-branch attention mechanism replaces standard multi-head self-attention to address its quadratic computational complexity.
2
Experiments on four public anomaly-detection benchmarks show that GTA outperforms alternative state-of-the-art methods.
3
GTA is a framework for multivariate IoT time-series anomaly detection that jointly learns sensor graphs, performs graph convolution, and models temporal dependencies with a Transformer-based architecture.
4
Influence Propagation convolution is introduced to model anomaly information flow between network nodes.
5
The framework automatically learns directed sensor connections using Gumbel-softmax sampling, avoiding reliance on known topology or predefined relationships.

multivariate time series generated by IoT sensor networks and infrastructures

anomalies and their spatiotemporal dependencies and inter-sensor influence relationships for detection

Publication Details
Publication Date
2021-07-27
Journal
Publisher
ISSN
Cited by
642
Access Type
Author Information
Authors
Zixuan Yuan
Xiuzhen Cheng
Dingshuo Chen
Xiao Zhang
Zekai Chen
Xiao Zhang
Explore further
Open the scid.ai AI chat with a ready-made request: it will find papers on a similar topic and help build a literature review.
Find similar papers in the chat →
Make a presentation
100%