Abstract Achieving a high‐quality interface between the topological insulator Bi 2 Se 3 and the high‐critical temperature superconductors YBa 2 Cu 3 O 7 (YBCO) remains challenging because of the rough surface morphology and elemental interdiffusion originating from YBCO. In previous studies, the Bi 2 Se 3 thin films directly deposited on YBCO exhibited poor crystallinity, which was significantly improved after a subsequent cracked‐Se treatment, although substantial elemental diffusion from YBCO still persisted. Here, we employ a two‐step growth technique consisting of an initial Bi thin‐film deposition followed by a cracked‐Se treatment and additionally introduce a thin Se buffer layer before the Bi deposition to effectively mitigate elemental diffusion. Structural and chemical characterizations demonstrate that the Se buffer layer effectively suppresses Cu diffusion, facilitates the formation of a uniform quintuple‐layer structure, and improves the crystallinity of the Bi 2 Se 3 thin films. Electrical measurements showed that YBCO with the Se buffer exhibits a higher superconducting transition temperature (~10 K) than the sample without the buffer (~4 K).

Abstract To address the challenges of traffic diversity and data imbalance in network intrusion detection, we propose GraphACGAN, a novel detection framework that integrates auxiliary classifier generative adversarial networks (ACGANs) with graph neural networks (GNNs). In this architecture, the GNN is embedded in the ACGAN discriminator to exploit the latent graph structures inherent to network traffic, thereby improving the model's capacity to distinguish between benign and malicious behaviors. Simultaneously, the ACGAN generator was leveraged to synthesize minority‐class attack traffic, effectively mitigating class imbalances and enhancing generalization. Comprehensive experiments conducted on three benchmark datasets, namely, NF‐BoT‐IoT‐v2, NF‐ToN‐IoT‐v2, and NF‐UNSW‐NB15‐v2, demonstrate that GraphACGAN consistently outperforms all baselines, including E‐GraphSAGE, GCN, ACGAN, LSTM, and KNN in terms of accuracy, precision, recall, and F 1 ‐score. An evaluation of resource‐limited computing platforms further demonstrates GraphACGAN's inference efficiency for real‐world deployment.

Abstract User reading status provides valuable insights into cognitive processes. Most reading classification methods rely on electrooculography (EOG). However, classifying EOG signals in uncontrolled environments poses challenges because of noise and limited data. To address this issue, methods based on self‐supervised learning or nested architectures have been proposed. However, their performance is often limited because they do not optimize the model in an end‐to‐end manner. Therefore, we propose an end‐to‐end network for two‐dimensional (2D) signal images generated from reshaped EOG signals. A 2D signal image was generated by reshaping EOG signals to incorporate both horizontal and vertical eye movements along with their magnitudes. We designed a ResNet‐based network to classify 2D signal images and introduced data augmentation techniques commonly used in image classification tasks. Our experiments, conducted using a publicly available dataset, evaluated various factors such as network structures, segmentation strategies, sampling rates, and sensor modalities. The results demonstrate that the proposed approach significantly improves classification accuracy compared with existing methods.

Abstract Texts are widely used in natural language processing. However, such applications are vulnerable to adversarial attacks. Existing research attempts to artificially add semantically meaningless word‐, character‐, or sentence‐level perturbations, which compromise the syntax and consistency of texts. However, they fail to ensure high‐quality outputs. Therefore, we propose an attack model for generating adversarial samples using policy gradients and a generative adversarial network. In our model, first, a Seq2Seq encoder is used to generate sentences, mapping discrete text data into continuous hidden space vectors and then transforming them into adversarial text samples. Second, to emphasize semantics, we compute the cosine similarity or BERT‐based semantic similarity between the original and adversarial texts for reward calculation. Finally, a policy gradient is applied to optimize the parameters. Experiments show that, while maintaining a semantic similarity above 0.8, our BERT‐based method reduces classification accuracy by 51.77% on the DBpedia dataset. Our cosine similarity‐based method requires only one‐third to one‐half the runtime of the baseline approach.

Abstract Edge computing deploys computing resources to the network edge to diminish task processing delays and power consumption. However, in traffic video surveillance systems, vehicular movement can lead to service interruptions. Moreover, the low credibility of the edge systems affects the success rate of edge offloading. To address these issues, we propose a secure task offloading scheme for traffic video surveillance. This scheme comprehensively evaluates trust values by integrating direct and indirect trust values, selecting nodes based on predefined trust thresholds, and achieving secure offloading and migration with short delays and energy consumption. To model the service migration problem, we adopted a Markov decision process‐based approach and employed a ‐learning algorithm to determine the optimal solution, thereby establishing an effective migration path. Simulation results demonstrate that the proposed scheme yields higher comprehensive trust values and improved reliability. Furthermore, while ensuring seamless task migration, the migration success rate exceeded 90%, with the energy consumption reduced by more than 9.8%.

Abstract The complexity of modern battlefields demands advanced military training systems that prepare forces for realistic scenarios. Even though traditional training methods are effective, they are costly and time consuming and are associated with safety risks. Virtual training systems offer a safer and more cost‐effective alternative; however, current solutions often compromise realism because of the need for multiple sensors and wearable devices that can diminish immersion. To address these limitations, we propose a marker‐based, adaptive, virtual military training system (MAVMTS) that enhances realism by using a minimal set of fiducial markers and multiview cameras to estimate the trainee's posture and weapon orientation without cumbersome wearables. The system integrates action recognition to generate responsive virtual adversaries, thereby creating dynamic and immersive training environments. MAVMTS reduces considerably the equipment burden and enhances the realism of virtual military training, thereby offering a more effective solution for preparing personnel for modern warfare.

Abstract We experimentally demonstrate 100 Gb/s bidirectional transmission over 40 km using a multi‐wavelength bidirectional optical sub‐assembly (BOSA) based on a single bidirectional multi‐wavelength Mux/Demux. The Mux/Demux consists of an optical zig‐zag glass block and thin film filters. Four wavelengths with 800 GHz spacing (that is, two wavelengths for each direction) within a 20‐nm band in the O‐band are utilized, and there is no four‐wave mixing penalty. The multi‐wavelength BOSA enables the bidirectional transmission of 2 × 50 Gb/s PAM4 signals over a 40‐km single‐fiber link. The multi‐wavelength BOSA with multi‐level modulation presents a suitable approach for future speed and transmission distance increases in 6G networks and data‐center interconnections.

Abstract The information interaction and positioning technology based on commercial Wi‐Fi plays a crucial role in expanding the applications for the Internet of Things and smart homes. Achieving high‐precision indoor positioning through the optimal deployment of Wi‐Fi access points remains a significant challenge. This study proposes two indoor positioning optimization algorithms based on the received signal strength indicator (RSSI) of commercial Wi‐Fi signals. The enhanced weighted centroid positioning (EWCP) algorithm introduces a novel weighting method that effectively leverages RSSI ranging errors to improve positioning accuracy and further minimize positioning errors. Meanwhile, the adaptive iterative weighted centroid positioning (AIWCP) algorithm incorporates a detailed weighting approach during the movement of the signal transmitter, providing specific movement indications and enhancing positioning performance. Both simulation and experimental results confirm the effectiveness of these two positioning methods.

Abstract Existing spatial feature recognition and layout methods primarily identify spatial components manually, which is time‐consuming and inefficient, and the constraint relationship between objects in space can be difficult to observe. Consequently, this study introduces an advanced spatial feature recognition and layout methodology employing enhanced CenterNet and LSTM frameworks, which is bifurcated into two major components—first, HCenterNet‐based feature recognition enhances feature extraction through an attention mechanism and feature fusion technology, refining the identification of small targets within complex background areas; second, a GA‐BiLSTM‐based spatial layout model uses a bidirectional LSTM network optimized with a genetic algorithm (GA), aimed at fine‐tuning the network parameters to yield more accurate spatial layouts. Experiments verified that compared with the CenterNet model, the recognition performance of the proposed HCenterNet‐DIoU model improved by 7.44%. Moreover, the GA‐BiLSTM model improved the overall layout accuracy by 10.08% compared with the LSTM model. Time cost analysis also confirmed that the proposed model could meet the real‐time requirements.