IET Cyber-Physical Systems: Theory and Applications最新文献

Dependency analysis can better help neural network to capture semantic features in sentences, so as to extract entity relation. Currently, hard pruning strategies and soft pruning strategies based on dependency tree structure coding have been proposed to balance beneficial additional information and adverse interference in extraction tasks. A new model based on graph convolutional networks, which uses a variety of representations describing dependency trees from different perspectives and combining these representations to obtain a better sentence representation for relation classification is proposed. A newly defined module is added, and this module uses the attention mechanism to capture deeper semantic features from the context representation as the global semantic features of the input text, thus helping the model to capture deeper semantic information at the sentence level for relational extraction tasks. In order to get more information about a given entity pair from the input sentence, the authors also model implicit co-references (references) to entities. This model can extract semantic features related to the relationship between entities from sentences to the maximum extent. The results show that the model in this paper achieves good results on SemEval2010-Task8 and KBP37 datasets.

Recent work has shown that deep reinforcement learning (DRL) is vulnerable to adversarial attacks, so that exploiting vulnerabilities in DRL systems through adversarial attack techniques has become a necessary prerequisite for building robust DRL systems. Compared to traditional deep learning systems, DRL systems are characterised by long sequential decisions rather than one-step decision, so attackers must perform multi-step attacks on them. To successfully attack a DRL system, the number of attacks must be minimised to avoid detecting by the victim agent and to ensure the effectiveness of the attack. Some selective attack methods proposed in recent researches, that is, attacking an agent at partial time steps, are not applicable to real-time attack scenarios, although they can avoid detecting by the victim agent. A real-time selective attack method that is applicable to environments with discrete action spaces is proposed. Firstly, the optimal attack threshold T for performing selective attacks in the environment Env is determined. Then, the observation states corresponding to when the value of the action preference function of the victim agent in multiple eposides exceeds the threshold T are added to the training set according to this threshold. Finally, a universal perturbation is generated based on this training set, and it is used to perform real-time selective attacks on the victim agent. Comparative experiments show that our attack method can perform real-time attacks while maintaining the attack effect and stealthiness.

Single-image dehazing is a highly challenging ill-posed task in the field of computer vision. To address this, a new image dehazing model with feature attention, named feature attention gated context aggregation network (FAGCA-Net), is proposed to tackle the issues of incomplete or over-dehazing caused by the original model's inability to handle non-uniform haze density distributions. A feature attention module that combines channel attention and spatial attention is introduced. Additionally, the authors propose a new extended attention convolutional block, which not only addresses the grid artefacts caused by the extended convolution but also provides added flexibility in handling different types of feature information. At the same time, in addition to the input image itself, incorporating the dark channel and edge channel of the image as the final input of the model is helpful for the model learning process. To demonstrate the robustness of the new model, it is applied to two completely different dehazing datasets, and it achieves significant dehazing performance improvement over the original model. Finally, to verify the effectiveness of the model in practical production processes, the authors apply it as an image preprocessing step to a set of UAV (Unmanned Aerial Vehicle) images of foreign objects. The result shows that the UAV images after being processed by FAGCA-Net for haze removal have a better impact on subsequent usage.

Cyberattacks on cyber-physical systems (CPS) have the potential to cause widespread disruption and affect the safety of millions of people. Machine learning can be an effective tool for detecting attacks on CPS, including the most stealthy types of attacks, known as covert channel attacks. In this study, the authors describe a novel hierarchical ensemble architecture for detecting covert channel attacks in CPS. Our proposed approach uses a combination of TCP payload entropy and network flows for feature engineering. Our approach achieves high detection performance, shortens the model training duration, and shows promise for effective detection of covert channel communications. This novel architecture closely mirrors the CPS attack stages in real-life, providing flexibility and adaptability in detecting new types of attacks.

The dimension of relevant text feature space and feature weight of substation main equipment defect information is high, so it is difficult to accurately select mining features. The Natural Language Processing (NLP) medium and short-term neural network model is used to realise the defect information text feature word segmentation in the log. After extracting the text features of defect information of main substation equipment with high categories to form the feature space; the TF-IDF algorithm is designed to calculate the importance weight of text keywords, judge the criticality of defect information text feature vocabulary, accurately locate defect information text features, and realise defect information text feature mining. Experiments show that the algorithm has high precision for specific word segmentation of massive substation main equipment log information.

The authors propose a novel object detection algorithm for identifying bird nests in medium voltage power line aerial images, which is crucial for ensuring the safe operation of the power grid. The algorithm utilises an improved Swin Transformer as the main feature extraction network of Fast R-CNN, further enhanced with a channel attention and modified binary self-attention mechanism to improve the feature representation ability. The proposed algorithm is evaluated on a newly constructed image dataset of medium voltage transmission lines containing bird nests, which are annotated and classified. Experimental results show that the proposed algorithm achieves satisfied accuracy and robustness in recognising bird nests compared to traditional algorithms.

An asymmetric semantic multi-level segmentation method based on depth learning is proposed in order to improve the precision and effect of semantic segmentation. A ‘content tree’ structure and an adjacency matrix are constructed to represent the parent-child relationship between each image sub region in a complete image. Through multiple combinations of spatial attention mechanism and channel attention mechanism, the similarity semantic features of the target object can be selectively aggregated, so as to enhance its feature expression and avoid the impact of significant objects. The asymmetric semantic segmentation model asymmetric pyramid feature convolutional network (APFCN) is constructed, and the path feature extraction and parameter adjustment are realised through APFCN. On the basis of APFCN network, a full convolution network is introduced for end-to-end image semantic segmentation. Combining the advantages of convolution network in extracting image features and the advantages of short-term and short-term memory network in solving long-term dependence, an end-to-end hybrid depth network is constructed for image semantic multi-level segmentation. The experimental results show that the mean intersection over Union value and mean pixel accuracy value are higher than that of the literature method, both of which are increased by more than 3%, and the segmentation effect is good.

In industrial environments, workers should wear workwear for safety considerations. For the same reason, smoking is also prohibited. Due to the supervision of monitoring devices, workers have reduced smoking behaviours and started wearing workwear. To meet the requirements for detecting these behaviours in real-time monitoring videos with high speed and accuracy, the authors proposed an improved YOLOv5 model with the Triplet Attention mechanism. This mechanism strengthens the connection between channel and spatial dimensions, focuses the network on important parts, and improves feature extraction. Compared to the original YOLOv5 model, the addition of the mechanism increases the parameters by only 0.04%. The recall rate of the YOLOv5 model is enhanced while its prediction speed is maintained with only a minimal increase in parameters. Experiment results show that, compared to the original model, the improved YOLOv5 has a recall rate of 78.8%, 91%, and 89.3% for detecting smoking behaviour, not wearing helmets, and inappropriate workwear, respectively.

Within the Industry 4.0 landscape, humans collaborate with cyber and physical elements to form human-cyber-physical systems (HCPS). These environments are increasingly complex and challenging workspaces due to increasing levels of automation and data availability. An effective system design requires suitable frameworks that consider human activities and needs whilst supporting overall system efficacy. Although several reviews of frameworks for technology were identified, none of these focused on the human in the system (moving towards Industry 5). The critical literature review presented provides a summary of HCPS frameworks, maps the considerations for a human in HCPS, and provides insight for future framework and system development. The challenges, recommendations, and areas for further research are discussed.

Online presence is becoming an important part of everyday's life and online communities may represent a significant source of engagement for the elderlies. Nevertheless, many may struggle to be online due to a lack of expertise, and a decentralised architecture may provide a solution by removing intermediaries, such as a webmaster, while not requiring expensive cloud solutions. However, issues concerning accessibility, security, and user experience have to be tackled. The paper focuses mainly on three issues: providing a human-readable domain, moderating content, and creating a reward system based on user reputation. An architecture is proposed based on Ethereum and Swarm. Smart contracts provide an automated set of rules to handle enterprise registration, content creation, and decision-making process, while Swarm serves both as distributed storage and the web host. Besides, in combination with Ethereum Name Service, Swarm provides a secure, distributed, and human-readable point of access to the web interface. The paper also describes an innovative two-token system where one token is meant to be a trustworthy reputation metre and the other is a spendable coin to get rewards. The final result is a fully decentralised, authenticated and moderated platform where users can aggregate and share their content presentations on the Internet.