Adversarial Network Research Articles

Low‐Frequency Reconstruction for Full Waveform Inversion by Unsupervised Learning

AbstractObtaining reliable low‐frequency seismic data is crucial for effectively reducing cycle‐skipping in full waveform inversion. However, acquiring high signal‐to‐noise ratio low‐frequency information from field data remains a challenge. An effective solution to mitigate cycle‐skipping is to utilize low‐frequency information synthesized by neural networks to obtain low‐wavenumber initial models. Previous attempts to reconstruct synthetic low‐frequency data using supervised learning methods have shown feasibility but were limited to training with synthetic data that required labeled information. In this study, we employed an unsupervised learning method, namely cycle‐consistent adversarial networks (CycleGAN), to reconstruct large‐scale‐feature related low‐frequency information based on the high‐frequency input data. Unlike supervised learning, CycleGAN allows the use of field data as input to train the network, which is more closely aligned with practical applications. Nevertheless, this approach presents challenges in terms of training complexity and potential output stability. To overcome these challenges, we reconstructed an appropriate target data set that combines high, medium, and low‐frequency components and incorporated additional loss functions to enhance the network's output performance. We conducted quantitative evaluations of the method's sensitivity to the target data set and its ability to handle low‐quality input data through numerical testing. The final results from field data testing confirmed the feasibility and effectiveness of the proposed method.

Advances in Medical Image Segmentation: A Comprehensive Review of Traditional, Deep Learning and Hybrid Approaches.

Medical image segmentation plays a critical role in accurate diagnosis and treatment planning, enabling precise analysis across a wide range of clinical tasks. This review begins by offering a comprehensive overview of traditional segmentation techniques, including thresholding, edge-based methods, region-based approaches, clustering, and graph-based segmentation. While these methods are computationally efficient and interpretable, they often face significant challenges when applied to complex, noisy, or variable medical images. The central focus of this review is the transformative impact of deep learning on medical image segmentation. We delve into prominent deep learning architectures such as Convolutional Neural Networks (CNNs), Fully Convolutional Networks (FCNs), U-Net, Recurrent Neural Networks (RNNs), Adversarial Networks (GANs), and Autoencoders (AEs). Each architecture is analyzed in terms of its structural foundation and specific application to medical image segmentation, illustrating how these models have enhanced segmentation accuracy across various clinical contexts. Finally, the review examines the integration of deep learning with traditional segmentation methods, addressing the limitations of both approaches. These hybrid strategies offer improved segmentation performance, particularly in challenging scenarios involving weak edges, noise, or inconsistent intensities. By synthesizing recent advancements, this review provides a detailed resource for researchers and practitioners, offering valuable insights into the current landscape and future directions of medical image segmentation.

Social Media Authentication and Combating Deepfakes using Semi-fragile Invisible Image Watermarking

With the significant advances in deep generative models for image and video synthesis, Deepfakes and manipulated media have raised severe societal concerns. Conventional machine learning classifiers for deepfake detection often fail to cope with evolving deepfake generation technology and are susceptible to adversarial attacks. Alternatively, invisible image watermarking is being researched as a proactive defense technique that allows media authentication by verifying an invisible secret message embedded in the image pixels. A handful of invisible image watermarking techniques introduced for media authentication have proven vulnerable to basic image processing operations and watermark removal attacks. In response, we have proposed a semi-fragile image watermarking technique that embeds an invisible secret message into real images for media authentication. Our proposed watermarking framework is designed to be fragile to facial manipulations or tampering while being robust to benign image-processing operations and watermark removal attacks. This is facilitated through a unique architecture of our proposed technique consisting of critic and adversarial networks that enforce high image quality and resiliency to watermark removal efforts, respectively, along with the backbone encoder-decoder and the discriminator networks. This allows images shared over the Internet to retain the verifiable watermark as long as facial manipulations or any other Deepfake modification technique is not applied. Thorough experimental investigations on SOTA facial Deepfake datasets demonstrate that our proposed model can embed a \(64\) -bit secret as an imperceptible image watermark that can be recovered with a high-bit recovery accuracy when benign image processing operations are applied while being non-recoverable when unseen Deepfake manipulations are applied. In addition, our proposed watermarking technique demonstrates high resilience to several white-box and black-box watermark removal attacks. Thus, obtaining state-of-the-art performance.

Cross-domain Intelligent Diagnostics for Rotating Machinery Using Domain Adaptive and Adversarial Networks

Cross-domain Intelligent Diagnostics for Rotating Machinery Using Domain Adaptive and Adversarial Networks

Multiscale feature-based robust secure diffusion estimation with noisy input over adversarial networks

This paper studies secure distributed estimation over wireless sensor networks in adversarial and noisy environments where a subset of sensors may be invaded by malicious attacks. We build a system model for the problem and prove that the bias-compensation method can resist noisy input but fails in attack defence. To alleviate the impact of attacks, a multiscale feature-based attack detection algorithm is proposed consisting of two main steps. Firstly, a binary discrimination mechanism is proposed to split neighbors of each node into two groups by their relative-state whose perception explores and exploits delicate element-wised features of attack vector. To reduce computational costs, the classifier is designed in a tree shape. Specifically, at each internal tree-node, a generalized correntropy based similarity metric function is developed and compared with an optimal threshold. Secondly, based on features of the classified groups, an absolute-state decider is presented to detect the trust neighbors and thus secure information sharing can be achieved. Simulation results reveal the effectiveness and robustness of our proposed method compared with some state-of-the-art algorithms under various attack forms.

Privacy Preservation of Large Language Models in the Metaverse Era: Research Frontiers, Categorical Comparisons, and Future Directions

ABSTRACTLarge language models (LLMs), with their billions to trillions of parameters, excel in natural language processing, machine translation, dialog systems, and text summarization. These capabilities are increasingly pivotal in the metaverse, where they can enhance virtual interactions and environments. However, their extensive use, particularly in the metaverse's immersive platforms, raises significant privacy concerns. This paper analyzes existing privacy issues in LLMs, vital for both traditional and metaverse applications, and examines protection techniques across the entire life cycle of these models, from training to user deployment. We delve into cryptography, embedding layer encoding, differential privacy and its variants, and adversarial networks, highlighting their relevance in the metaverse context. Specifically, we explore technologies like homomorphic encryption and secure multiparty computation, which are essential for metaverse security. Our discussion on Gaussian differential privacy, Renyi differential privacy, Edgeworth accounting, and the generation of adversarial samples and loss functions emphasizes their importance in the metaverse's dynamic and interactive environments. Lastly, the paper discusses the current research status and future challenges in the security of LLMs within and beyond the metaverse, emphasizing urgent problems and potential areas for exploration.

An adversarial network used for drift correction in electronic nose

The electronic nose (E-nose) is an important means for machines to sense odor signals, but it faces a serious problem that limits its development: sensor drift. Drift refers to the long-time and irregular changes in the response signal of the gas sensor to the same analyte, and it will bring serious performance degradation to the E-nose. To reduce the effect of drift, we proposed an entropy minimization model based on adversarial networks (EMAD) which reduced the distribution difference between data with and without drift through adversarial training. EMAD consists of a generator and a classifier. To simplify the structure and improve the prediction diversity, the classifier is reused as a discriminator for adversarial training while learning the domain invariant features from the source domain. EMAD reduced the distribution difference between original data and drift data while preserving the features for classification. Experiment results on the public dataset showed that EMAD effectively eliminated the performance degradation caused by drift and achieved the highest classification accuracy compared with other advanced methods.

Investigation on situation awareness model of unmanned aerial vehicle groups communication network based on adversarial network

With the widespread application and development of unmanned aerial vehicle (UAV) technology, ensuring the security and stability of UAV swarm communication networks has become crucial. Given the diverse forms of interference and attacks in current networks, this poses a serious threat to the normal operation of UAV swarm communication. Therefore, how to accurately identify and effectively counter these network threats has become the focus of research. This study comprehensively evaluates the core technology of UAV swarm communication network situational awareness and constructs a situational awareness model based on adversarial networks. The model utilizes adversarial network technology and combines data collection and processing to design four experiments to comprehensively evaluate the performance of the model in different scenarios. The experimental results show that as the amount of data gradually increases, the performance of the model also improves. When processing 100, 1000, and 10,000 data points, the model achieved accuracies of 0.955, 0.962, and 0.982, respectively. Furthermore, the experimental results also indicate that effective noise suppression measures can significantly improve the accuracy and stability of the situational awareness model. Additionally, it is noted that different model structures will affect training duration, accuracy, and stability. Although increasing network scale may lead to increased computational complexity and latency, its accuracy is correspondingly improved. The adversarial network-based situational awareness model proposed in this study can accurately identify and effectively counter interference and attacks in UAV swarm communication networks, thereby providing solid protection for the collaborative combat and information sharing of UAV swarms.

Class conditioned text generation with style attention mechanism for embracing diversity

In the field of artificial intelligence and natural language processing (NLP), natural language generation (NLG) has significantly advanced. Its primary aim is to automatically generate text in a manner resembling human language. Traditional text generation has mainly focused on binary style transfers, limiting the scope to simple transformations between positive and negative tones or between modern and ancient styles. However, accommodating style diversity in real scenarios presents greater complexity and demand. Existing methods usually fail to capture the richness of diverse styles, hindering their utility in practical applications. To address these limitations, we propose a multi-class conditioned text generation model. We overcome previous constraints by utilizing a transformer-based decoder equipped with adversarial networks and style-attention mechanisms to model various styles in multi-class text. According to our experimental results, the proposed model achieved better performance compared to the alternatives on multi-class text generation tasks in terms of diversity while it preserves fluency. We expect that our study will help researchers not only train their models but also build simulated multi-class text datasets for further research.

A Two-Stage Multi-Target Domain Adaptation Framework for Prediction of Key Performance Indicators Based on Adversarial Network

A Two-Stage Multi-Target Domain Adaptation Framework for Prediction of Key Performance Indicators Based on Adversarial Network

A novel decomposition and hybrid transfer learning-based method for multi-step cutterhead torque prediction of shield machine

As a core operating parameter of shield machine, the high-precision multi-step prediction of cutterhead torque can effectively help the driver adjust the operating parameters in advance to ensure the safe operation of shield machine. Due to data distribution discrepancy, it is challenging to accurately predict the cutterhead torque under different geological conditions. Currently, the existing methods cannot precisely predict cutterhead torque in cross domain. In order to accurately predict the cutterhead torque under different geological environments, we propose a novel decomposition and hybrid transfer learning-based method for multi-step cutterhead torque transfer prediction. First, original cutterhead torque signal is decomposed into some sub-signals by variational mode decomposition (VMD), which can effectively reduce the complexity. Then hybrid transfer learning-based method (HTLM) is established to predict each sub-signal and add the prediction results to realize multi-step cutterhead torque prediction in cross domain. HTLM has three network branches. The first one is that multi-layer GRU network is used to extract time-varying characteristics of source domain data. The second is that the automatic encoder is used to extract the deep common features between source domain and target domain. The third is the adversarial network and the maximum mean discrepancy (MMD) metric for domain adaptation. Then, the extracted features are integrated to predict the cutterhead torque in cross domain. To verify the proposed VMD-HTLM, several transfer prediction experiments were carried out. From the first step to fifth step transfer prediction, the proposed method achieved higher prediction accuracy than existing methods. Especially in the fifth step, the accuracy of proposed method is on average 7.09 % higher than existing methods, the RMSE and the MAE decreased on average 40.00 % and 39.92 %, respectively. Furthermore, the average accuracy of proposed method is 93.79 %, 93.58 %, 93.55 %, 92.25 % and 91.79 % from first step to fifth step, respectively. Hence, the proposed VMD-HTLM can accurately predict the cutterhead torque under different geological conditions.

Adversarial Robust Safeguard for Evading Deep Facial Manipulation

The non-consensual exploitation of facial manipulation has emerged as a pressing societal concern. In tandem with the identification of such fake content, recent research endeavors have advocated countering manipulation techniques through proactive interventions, specifically the incorporation of adversarial noise to impede the manipulation in advance. Nevertheless, with insufficient consideration of robustness, we show that current methods falter in providing protection after simple perturbations, e.g., blur. In addition, traditional optimization-based methods face limitations in scalability as they struggle to accommodate the substantial expansion of data volume, a consequence of the time-intensive iterative pipeline. To solve these challenges, we propose a learning-based model, Adversarial Robust Safeguard (ARS), to generate desirable protection noise in a single forward process, concurrently exhibiting a heightened resistance against prevalent perturbations. Specifically, our method involves a two-way protection design, characterized by a basic protection component responsible for generating efficacious noise features, coupled with robust protection for further enhancement. In robust protection, we first fuse image features with spatially duplicated noise embedding, thereby accounting for inherent information redundancy. Subsequently, a combination comprising a differentiable perturbation module and an adversarial network is devised to simulate potential information degradation during the training process. To evaluate it, we conduct experiments on four manipulation methods and compare recent works comprehensively. The results of our method exhibit good visual effects with pronounced robustness against varied perturbations at different levels.

Identification of Small Goods under the Adversarial Network

Identification of Small Goods under the Adversarial Network

Convergent and Divergent Corporate Social Responsibility in South Korea: Collaborative and Adversarial NGO-Corporate Networks

Background: The differences between NGO networks for two distinct types of CSR practices are underexplored: convergent CSR, which pertains to the global standards embraced by both the local and global institutions, and divergent CSR, which is framed primarily by local economic, political and social conditions. Purpose: Grounded in institutional and network theory, the study explores the significance of three forms of network centrality in different types of CSR (convergent/divergent) and varying modes of interaction (collaborative/adversarial) across global versus local NGOs in South Korea, a state-led market economy. Research Design: The study conducts network analyses and descriptive analysis of NGOs’ collaborative and adversarial networks with corporations, in relation to their engagement in convergent and divergent CSR. Study Sample: The NGO/corporate network dataset consisted of 2073 nodes and 4158 edges (ties). The sample of CSR practices consisted of a total of 8715 instances of convergent CSR practices and 396 instances of divergent CSR practices. Data Collection and/or Analysis: A total of 260 reports from 52 South Korean corporations and a total of 430 reports from 78 NGOs in South Korea were used to develop a corpus of corporate/NGO network dataset (collaborative and adversarial) and the dataset of the type of CSR practices they engaged in (convergent/divergent). Then degree, eigenvector and betweenness centralities of the 78 NGOs were computed within the NGO-corporate network. The relationships between the three centralities and CSR types (convergent/divergent) were found through standard regression analyses and descriptive analyses. Results: The findings suggest that when engaging in convergent CSR, as opposed to divergent CSR, NGOs would benefit the most from developing collaborative ties to central others in their NGO-corporate network (eigenvector centrality). A descriptive analysis of the findings suggests that adversarial divergent CSR practices are primarily reported by a potentially isolated group of local NGOs. Conclusions: CSR practices develop in multiple forms within a national institution, rather than simply converging with the universal norms. They form under the communicative pressures of different types of global and local institutional actors, through different network positions (centralities) and nature of relationships (collaborative and adversarial).

ResUNet involved generative adversarial network-based topology optimization for design of 2D microstructure with extreme material properties

Topology optimization is one of the most common methods for design of material distribution in mechanical metamaterials, but resulting in expensive computational cost due to iterative simulation of finite element method. In this work, a novel deep learning-based topology optimization method is proposed to design mechanical microstructure efficiently for metamaterials with extreme material properties, such as maximum bulk modulus, maximum shear modulus, or negative Poisson’s ratio. Large numbers of microstructures with various configurations are first simulated by modified solid isotropic material with penalization (SIMP), to construct the microstructure data set. Subsequently, the ResUNet involved generative and adversarial network (ResUNet-GAN) is developed for high-dimensional mapping between optimization parameters and corresponding microstructures to improve the design accuracy of ResUNet. By given optimization parameters, the well-trained ResUNet-GAN is successfully applied to the microstructure design of metamaterials with different optimization objectives under proper configurations. According to the simulation results, the proposed ResUNet-GAN-based topology optimization not only significantly reduces the computational duration for the optimization process, but also improves the structure precise and mechanical performance.

Nuclei segmentation using attention aware and adversarial networks

Accurate segmentation of nuclei plays a critical role in pathology since assessments and diagnoses are mainly based on the recognition, measurement, and counting of nuclei. However, in digital pathology, automated nucleus segmentation is a challenging issue because of various factors such as color inconsistency in stained images, unclear nucleus boundaries, low contrast between background and nuclei, different shapes and sizes of nuclei, and intensity inhomogeneity not only inside nuclei but also across them and the background. In this work, an efficient method has been developed for nuclei segmentation. Its efficiency has been achieved through tissue and chemical invariant normalization, feature extraction with dense convolution layers, merging of local and global features, mask generation, and adversarial learning. For fair comparisons, state-of-the-art nuclei segmentations have been employed to the same data sets, and their performances have been evaluated using the same metrics. This paper’s main contributions are threefold: (i) Introducing a novel technique for nuclei segmentation using an adversarial network and a hybrid attention-aware network. (ii) Presenting the effective merging of global and local features to enhance pattern recognition, and the utilization of efficient hybrid attention blocks for extracting desired global information and improving relationships between feature regions at different locations. (iii) Presenting experimental results showing that the proposed technique accomplishes nuclei detection and extraction with higher accuracy (a minimum improvement of 3.7%) than other recent methods. Also, each stage provides considerable contributions to the segmentation performance. Particularly, the hybrid attention-aware network has improved the performance by 4.2% according to the dice coefficient.

Systematic Face Pareidolia Generation Method Using Cycle‐Consistent Adversarial Networks

Pareidolia is a psychological tendency of perceiving a face in non‐face stimulus. As a majority of people globally experience this tendency, it has been extensively studied and measured in terms of tendencies, such as frequencies. However, no study has investigated the systematic manipulation of stimulus owing to the lack of a systematic image‐generation method. Therefore, herein, we generated face pareidolia stimuli using a face data set with annotated data. We employed cycle‐consistent adversarial networks (CycleGAN), an image‐to‐image‐style translation framework, to generate stimuli for translating natural‐image styles from face images. We manipulated the weight of the cycle‐consistency loss in the CycleGAN via an experiment to evaluate the image generated using the CycleGAN. Thus, we found that the weight value of the evaluation experiment correlated with the pareidolia‐inducing power when the preprocessing of the face data set was applied to the blur process. As a result, we achieved to systematically generate pareidolia stimuli. © 2024 The Authors. IEEJ Transactions on Electrical and Electronic Engineering published by Institute of Electrical Engineer of Japan and Wiley Periodicals LLC.

Video-based working condition recognition of fused magnesium furnace with stochastic configuration networks

Working condition recognition of a fused magnesium furnace (FMF) suffers from unbalanced under-burning condition samples, inconsistent quality of training samples and difficulty in characterizing dynamic production processes with images. This paper presents a novel approach to detect the under-burning working condition in FMF based on a 3D-cycle-generative adversarial network (3D-cycle-GAN) and Video Swin Transformer-Stochastic Configuration Networks (Video Swin-SCNs). Firstly, to resolve the temporal discontinuity defect caused by Cycle-GAN through the visual appearance of video composite frames, we construct a motion consistency-based 3D-Cycle-GAN model that considers the visual appearance and temporal continuity constraints of unpaired video transitions and is designed to generate video samples of under-burning working conditions. Secondly, a reinforcement learning approach is used to assess the value of the video quality and to filter out possible low-quality samples generated. Finally, the local attention is extended from the spatial domain to the spatial-temporal domain to solve the difficulty in characterizing the dynamic production process with the static images. The spatial-temporal features from the Video Swin Transformer are fed into SCNs to classify the working conditions. The experiment results indicate the effectiveness and feasibility of the proposed method.

Adversarial Modality Alignment Network for Cross-Modal Molecule Retrieval

The cross-modal molecule retrieval (Text2Mol) task aims to bridge the semantic gap between molecules and natural language descriptions. A solution to this non-trivial problem relies on graph convolutional network (GCN) and cross-modal attention with contrastive learning for reasonable results. However, there exist the following issues: 1) the cross-modal attention mechanism is only in favor of text representations and can not provide helpful information for molecule representations. 2) the GCN-based molecule encoder ignores edge features and the importance of various substructures of a molecule. 3) the retrieval learning loss function is rather simplistic. This paper further investigates the Text2Mol problem and proposes a novel Adversarial Modality Alignment Network (AMAN)-based method to sufficiently learn both description and molecule information. Our method utilizes a SciBERT as a text encoder and a graph transformer network as a molecule encoder to generate multimodal representations. Then an adversarial network is used to align these modalities interactively. Meanwhile, a triplet loss function is leveraged to perform retrieval learning and further enhance the modality alignment. Experiments on the ChEBI-20 dataset show the effectiveness of our AMAN method compared with baselines.

Prior Knowledge-guided Triple-Domain Transformer-GAN for Direct PET Reconstruction from Low-Count Sinograms.

To obtain high-quality positron emission tomography (PET) images while minimizing radiation exposure, numerous methods have been dedicated to acquiring standard-count PET (SPET) from low-count PET (LPET). However, current methods have failed to take full advantage of the different emphasized information from multiple domains, i.e., the sinogram, image, and frequency domains, resulting in the loss of crucial details. Meanwhile, they overlook the unique inner-structure of the sinograms, thereby failing to fully capture its structural characteristics and relationships. To alleviate these problems, in this paper, we proposed a prior knowledge-guided transformer-GAN that unites triple domains of sinogram, image, and frequency to directly reconstruct SPET images from LPET sinograms, namely PK-TriDo. Our PK-TriDo consists of a Sinogram Inner-Structure-based Denoising Transformer (SISD-Former) to denoise the input LPET sinogram, a Frequency-adapted Image Reconstruction Transformer (FaIR-Former) to reconstruct high-quality SPET images from the denoised sinograms guided by the image domain prior knowledge, and an Adversarial Network (AdvNet) to further enhance the reconstruction quality via adversarial training. Specifically tailored for the PET imaging mechanism, we injected a sinogram embedding module that partitions the sinograms by rows and columns to obtain 1D sequences of angles and distances to faithfully preserve the inner-structure of the sinograms. Moreover, to mitigate high-frequency distortions and enhance reconstruction details, we integrated global-local frequency parsers (GLFPs) into FaIR-Former to calibrate the distributions and proportions of different frequency bands, thus compelling the network to preserve high-frequency details. Evaluations on three datasets with different dose levels and imaging scenarios demonstrated that our PK-TriDo outperforms the state-of-the-art methods.