Anomaly detection in multivariate time series (MTS) has been applied to various areas. Recent studies for detecting anomalies in high-dimensional data have yielded promising results. However, these methods are incapable of explicitly dealing with the complex contextual information that exists between features. In this article, we present a novel unsupervised anomaly detection framework for MTS. We model the complex relationships of MTS using graph attention networks from the perspectives of time and features, respectively. Furthermore, our framework employs masked autoregressive flow for density estimation, which is then treated as an anomaly score, to identify anomalies. Extensive experiments show that our model outperforms baseline approaches in terms of accuracy on three publicly available data sets and accurately captures temporal and interfeature relationships.

Improvements in the emergency response efficiency and management level of power construction sites are conducive to reducing the construction safety risk of power infrastructure projects and then achieving the sustainability of construction site safety. Therefore, this paper proposes an emergency decision-making method for electric power personal accidents, which applies ontology and case-based reasoning to electric power emergency decision making. Firstly, ontology technology is used to structurally represent power accident case knowledge and clarify concepts and their relationships. Then, a power accident knowledge ontology hierarchy is designed, and a powerful personal accident case library is established. Secondly, by calculating cases’ conceptual similarity, attribute similarity, and structural similarity, a global power accident case similarity calculation method is proposed, and case matching is performed based on the calculation results to achieve case knowledge retrieval and reuse. Finally, the results of the example-based study show that the method effectively achieves the accurate retrieval of electric power accident cases, improves the efficiency of the emergency decision response to electric power construction site accidents, and then provides support for emergency decision making for electric power construction site accidents.

Unknown threats have caused severe damage in critical infrastructures. To solve this issue, the graph-based methods have been proposed because of their ability for learning complex interaction patterns of network entities with discrete graph snapshots. However, such methods are challenged by the computer networking model characterized by the natural continuous-time dynamic heterogeneous graph (CDHG). In this paper, we propose a CDHG-based graph neural network model, namely, CDHGN, for unknown threat detection. It first constructs the CDHG using interaction relationships among network entities extracted from various log records. Then, it trains the detection model based on a heterogeneous attention network and performs streaming detection for live online network events. We implement a prototype and conduct extensive experiments on a comprehensive cybersecurity dataset with more than nine million records. Experimental result shows that the proposed method can achieve superior detection performance than the state-of-the-art methods.

This paper adopts a low power consumption ARM Hisilicon mobile processing platform and OV4689 camera, combined with a new skeleton extraction based on distance transform algorithm and the improved Hough algorithm for multi meters real-time reading. The design and implementation of the device were completed. Experimental results shows that The average error of measurement was 0.005MPa, and the average reading time was 5s. The device had good stability and high accuracy which meets the needs of practical application.