Graph neural network‐enhanced auxiliary classifier generative adversarial network framework for robust intrusion detection

Machine learning is applied widely in the field of intrusion detection at present, the existing intrusion detection algorithm is relatively mature, but how to solve the problem of unbalanced sample data still need further research. Aiming at the problem of detection accuracy and efficiency caused by sample data imbalance in the process of intrusion detection, this paper proposes an intrusion detection method based on the fusion of Auxiliary Classification Adversarial Network (ACGAN) and Graph Neural Network (GNN) (ACGAN-GNN). Firstly, ACGAN is used to expand minority samples in data preprocessing to optimize the dataset, and then the improved heterogeneous graph neural network algorithm is used to model the sample flow relationship in the classification process, so as to improve the detection robustness of minority attack samples and unknown attack samples. The model is evaluated by CICIDS2017 dataset. Compared with similar algorithms, ACGAN-GNN not only has better performance in terms of accuracy, precision, recall and F1-score, but also has higher accuracy against minority or unknown attack types.

The imbalance of dissolved gas analysis (DGA) data will lead to over-fitting, weak generalization and poor recognition performance for fault diagnosis models based on deep learning. To handle this problem, a novel transformer fault diagnosis method based on improved auxiliary classifier generative adversarial network (ACGAN) under imbalanced data is proposed in this paper, which meets both the requirements of balancing DGA data and supplying accurate diagnosis results. The generator combines one-dimensional convolutional neural networks (1D-CNN) and long short-term memories (LSTM), which can deeply extract the features from DGA samples and be greatly beneficial to ACGAN’s data balancing and fault diagnosis. The discriminator adopts multilayer perceptron networks (MLP), which prevents the discriminator from losing important features of DGA data when the network is too complex and the number of layers is too large. The experimental results suggest that the presented approach can effectively improve the adverse effects of DGA data imbalance on the deep learning models, enhance fault diagnosis performance and supply desirable diagnosis accuracy up to 99.46%. Furthermore, the comparison results indicate the fault diagnosis performance of the proposed approach is superior to that of other conventional methods. Therefore, the method presented in this study has excellent and reliable fault diagnosis performance for various unbalanced datasets. In addition, the proposed approach can also solve the problems of insufficient and imbalanced fault data in other practical application fields.

Network-based Intrusion Detection(NID) is an effective means to deal with network attacks. NID is able to detect different types of network attacks by analyzing network traffic. However, in the real world, network traffic contains majority and minority class attacks as well as a large number of normal traffic samples. The imbalance in the number of training samples of various types of network traffic makes network intrusion detection very poor. Due to the lack of training samples, traditional NID can't learn the characteristics of minority class attacks, which leads to the failure of NID to detect minority class attacks. Therefore, in order to solve the problem brought by imbalanced data, we propose a network intrusion detection algorithm based on the sampling and improved One-vs-All(OVA) technique. The dataset is balanced by downsampling the majority class data based on K-means clustering and oversampling the minority class data based on Auxiliary Classifier Generative Adversarial Network(ACGAN), improve classification accuracy through OVA-based model training and testing. We conduct validation experiments on the NSL-KDD dataset, and the experimental results show that the proposed method achieves excellent results in terms of Accuracy, Precision, Recall and F1-score. Compared with existing state-of-the-art methods, the proposed method not only achieves excellent detection performance with low false positive rate, but also addresses the learning problem of imbalanced data more effectively.

Defective wafer pattern recognition is important for quality control and yield enhancement in semiconductor fabrication systems. The collected wafer maps are usually imbalanced, which may degrade the performance of classifier. In this paper, a focal auxiliary classifier generative adversarial network (FAC-GAN) for defective wafer pattern recognition with imbalanced data is proposed. FAC-GAN is composed of AC-GAN modified with focal loss for imbalanced data generation and deep neural network. The performance of the proposed FAC-GAN is measured on real-world wafer map dataset “WM-811k” and it outperforms SVM and CNN.

Deep learning-based methods have been well ap-plied in the field of fault diagnosis when enough data can be collected. However, the available fault samples are very limited in real-industrial scenarios. In this work, an enhanced auxiliary classifier generative adversarial network(EACGAN) based data generating techniques are proposed to solve the imbalanced fault diagnosis problem. We optimize the original auxiliary classifier generative adversarial network(ACGAN) in two ways. Firstly, the boundary seeking loss is adopted to stabilize the training of ACGANs. Secondly, the cost sensitive learning is introduced to make ACGANs more suitable to solve imbalanced problems. The validity test of the proposed method is carried out using the Tennessee Eastman (TE) dataset. The experimental results reveal that the proposed enhanced auxiliary classifier generative adversarial network(EACGAN) can achieve better performance than existing GAN-based methods.

TRA-ACGAN: A motor bearing fault diagnosis model based on an auxiliary classifier generative adversarial network and transformer network

In acoustic event detection, the training data size of some acoustic events is often small and imbalanced. To deal with this, this paper proposes generating the virtual training data categorically using the auxiliary classifier generative adversarial networks. Soft labels of acoustic events are first calculated to represent the acoustic event localization information. The closer the current frame is to the middle of the manually labeled acoustic event, the higher the soft label will be, which makes the soft labels positively correlated with the acoustic event localization. Then, the acoustic event class and the quantized soft labels are used as the input condition to the auxiliary classifier generative adversarial networks to generate an arbitrary number of training samples. Experimental results on the TUT Sound Event 2016 under the home environment and TUT Sound Event 2017 under the street environment demonstrate the improved performance of the proposed technique compared to existing acoustic event detection systems.

Imbalanced data classification: A KNN and generative adversarial networks-based hybrid approach for intrusion detection

Diagnosing faults in wheelset bearings is critical for train safety. The main challenge is that only a limited amount of fault sample data can be obtained during high-speed train operations. This scarcity of samples impacts the training and accuracy of deep learning models for wheelset bearing fault diagnosis. Studies show that the Auxiliary Classifier Generative Adversarial Network (ACGAN) demonstrates promising performance in addressing this issue. However, existing ACGAN models have drawbacks such as complexity, high computational expenses, mode collapse, and vanishing gradients. Aiming to address these issues, this paper presents the Transformer and Auxiliary Classifier Generative Adversarial Network (TACGAN), which increases the diversity, complexity and entropy of generated samples, and maximizes the entropy of the generated samples. The transformer network replaces traditional convolutional neural networks (CNNs), avoiding iterative and convolutional structures, thereby reducing computational expenses. Moreover, an independent classifier is integrated to prevent the coupling problem, where the discriminator is simultaneously identified and classified in the ACGAN. Finally, the Wasserstein distance is employed in the loss function to mitigate mode collapse and vanishing gradients. Experimental results using the train wheelset bearing datasets demonstrate the accuracy and effectiveness of the TACGAN.

Deep learning for robust detection of interictal epileptiform discharges

Recently, the Generative Adversarial Network (GAN) has attracted much attention in a large range of applications, such as image classification, plausible image generation and image completion. Lots of GAN variants have been proposed. This paper focused on two popular GAN variants, including GAN and Auxiliary Classifier Generative Adversarial Network (ACGAN) and made a comparison between them. The experiment on CIFAR-10 and anima dataset shows that ACGAN can perform better than original one with better accuracy. Furthermore, we find ACGAN has a worse inception score (IS), which indicates that ACGAN is not capable of generating realistic pictures.

Most of the existing laser welding process monitoring technologies focus on the detection of post-engineering defects, but in the mass production of electronic equipment, such as laser welding metal plates, the real-time identification of defect detection has more important practical significance. The data set of laser welding process is often difficult to build and there is not enough experimental data, which hinder the applications of the data-driven laser welding defect detection method. In this paper, an intelligent welding defect diagnosis method based on auxiliary classifier generative adversarial networks (ACGAN) has been proposed. Firstly, a ten-class dataset consisting of 6467 samples, was constructed, which originate from the optical and thermal sensory parameters in the welding process. A new structured ACGAN network model is proposed to generate fake data similar to the true defect feature distributions. In addition, in order to make the difference between different defects categories more obvious after data expansion, a data filtering and data purification scheme was proposed based on ensemble learning and an SVM (support vector machine), which is used to filter the bad generated data. In the experiments, the classification accuracy can reach 96.83% and 85.13%, for the CNN (convolutional neural network) algorithm model and ACGAN model, respectively. However, the accuracy can further improve to 97.86% and 98.37% for the fusion models of ACGAN-CNN and ACGAN-SVM-CNN models, respectively. The results show that ACGAN can not only be used as an algorithm model for classification, but also be used to achieve superior real-time classification and recognition through data enhancement and multi-model fusion.

Low‐magnitude earthquakes occur far more frequently than major quakes and often go unnoticed by the public. These tremors rarely cause any damage, yet they play an important role in advancing our understanding of Earth's seismicity. Accurate detection of low‐magnitude earthquakes is crucial to develop complete earthquake catalogs and improve seismic hazard forecasting models. However, conventional detection algorithms such as the short‐time‐average/long‐time‐average (STA/LTA) method struggle to identify these events because of their inherently low signal‐to‐noise ratio (SNR). Additionally, lack of labeled waveforms for low‐magnitude earthquakes further complicates the training of effective deep‐learning models. In this study, we use an Auxiliary Classifier Generative Adversarial Network (AC‐GAN) to produce synthetic yet realistic three‐component waveforms of low‐magnitude earthquakes. The AC‐GAN is trained on fixed‐length (60‐s) waveform segments conditioned by predefined SNR classes. All selected events have magnitudes lower than 3 and are categorized into 10 distinct SNR classes. Our results indicate that the AC‐GAN model generates realistic three‐component waveforms that effectively capture essential characteristics of real seismic signals. To evaluate the quality of these synthetic waveforms, we employ both quantitative and qualitative assessments. Quantitative analysis using Pearson's correlation coefficient yield relatively low correlations (ranging from 0.01 to 0.04); however, correlation values noticeably improve as SNR increases. Qualitatively, a user‐based visual inspection experiment demonstrate remarkable similarities in general seismic features between the synthetic and authentic waveforms. We also test their effectiveness for data augmentation in binary deep‐learning classifier designed for detecting low‐magnitude earthquakes. Our result show improved classification performance with the addition of synthetic data.

To address the class imbalance issue in network intrusion detection, which degrades performance of intrusion detection models, this paper proposes a novel generative model called VAE-WACGAN to generate minority class samples and balance the dataset. This model extends the Variational Autoencoder Generative Adversarial Network (VAEGAN) by integrating key features from the Auxiliary Classifier Generative Adversarial Network (ACGAN) and the Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP). These enhancements significantly improve both the quality of generated samples and the stability of the training process. By utilizing the VAE-WACGAN model to oversample anomalous data, more realistic synthetic anomalies that closely mirror the actual network traffic distribution can be generated. This approach effectively balances the network traffic dataset and enhances the overall performance of the intrusion detection model. Experimental validation was conducted using two widely utilized intrusion detection datasets, UNSW-NB15 and CIC-IDS2017. The results demonstrate that the VAE-WACGAN method effectively enhances the performance metrics of the intrusion detection model. Furthermore, the VAE-WACGAN-based intrusion detection approach surpasses several other advanced methods, underscoring its effectiveness in tackling network security challenges.

Classification of Imbalanced Electrocardiosignal Data using Convolutional Neural Network