Defending Network Intrusion Detection Systems Based on Graph Neural Networks Against Structural Adversarial Attacks
本文提出一种基于对抗训练的防御框架，通过生成对抗样本来增强图神经网络在网络安全入侵检测系统中对结构对抗攻击的鲁棒性。
0 citationsRead paper
Selected work
Representative Papers
A Tree-Structured Approach for Phishing Template and Attacker Attribution Analysis
This study addresses the limitations of existing phishing defenses in detecting template reuse and coordinated attacks by proposing a DOM tree-based fingerprinting approach coupled with hierarchical Jaccard distance metrics. By integrating HTML tag content augmentation with unsupervised clustering algorithms, the method enables precise extraction and visual analysis of webpage structural features. This approach effectively uncovers latent similarities among phishing sites, significantly enhancing the detection of emerging and zero-day templates. Furthermore, it provides robust technical support for coordinated threat analysis, thereby bridging critical gaps in traditional defense systems regarding structured correlation analysis. Collectively, this work advances phishing detection capabilities through rigorous structural feature engineering and similarity measurement.
0 citationsRead paper
A Machine Learning Framework for Real-Time Personalized Ergonomic Pose Analysis
Traditional single-view cameras struggle to deliver comprehensive, real-time ergonomic posture assessment due to fixed viewpoints and occlusion. This work proposes a real-time analysis system that fuses multi-view 3D point clouds with 2D pose estimation, uniquely leveraging user-annotated personalized samples to train a deep learning classifier and enabling continuous inference of ergonomic postures on streaming data. By transcending the limitations of a single viewpoint, the method effectively handles occluded scenarios and achieves high-accuracy skeletal labeling and posture recognition in load-carrying tasks, demonstrating its practicality and scalability for workplace health and safety monitoring.
0 citationsRead paper
Attention-based multiple instance learning for predominant growth pattern prediction in lung adenocarcinoma wsi using foundation models
This study addresses the challenge of lung adenocarcinoma grading, which relies on accurate identification of the predominant growth pattern but typically demands costly, slide-level fine-grained annotations. To mitigate this requirement, the authors propose an attention-based multiple instance learning (MIL) framework that operates solely with whole-slide image-level labels. The approach leverages a pretrained foundation model in computational pathology—such as Prov-GigaPath—as a patch encoder and employs an attention mechanism to aggregate global features for growth pattern prediction. This work represents the first integration of a pathology foundation model with attention-based MIL, substantially reducing dependence on pixel-level annotations while enhancing prediction robustness. Experimental results demonstrate that fine-tuned Prov-GigaPath within the ABMIL framework achieves state-of-the-art performance (Cohen’s κ = 0.699), significantly outperforming conventional aggregation baselines.
0 citationsRead paper
Identification and Anonymization of Named Entities in Unstructured Information Sources for Use in Social Engineering Detection
This study addresses the challenge of balancing data availability for cybercrime analysis with privacy protection under regulations such as the GDPR by proposing an end-to-end multimodal data processing pipeline. The pipeline integrates speech enhancement, high-precision named entity recognition (NER), and structure-preserving anonymization techniques to construct a compliant dataset from Telegram-collected text, audio, and images. Experimental results demonstrate that the Parakeet model achieves optimal speech transcription performance, while the proposed Transformer-based NER approach attains the highest F1 score in identifying sensitive information. Furthermore, the anonymized data retains essential semantic structures while satisfying regulatory compliance requirements, thereby effectively supporting research on social engineering attack detection.
0 citationsRead paper
Recent publications
Latest Papers
Defending Network Intrusion Detection Systems Based on Graph Neural Networks Against Structural Adversarial Attacks
本文提出一种基于对抗训练的防御框架，通过生成对抗样本来增强图神经网络在网络安全入侵检测系统中对结构对抗攻击的鲁棒性。
0 citationsRead paper
A Tree-Structured Approach for Phishing Template and Attacker Attribution Analysis
This study addresses the limitations of existing phishing defenses in detecting template reuse and coordinated attacks by proposing a DOM tree-based fingerprinting approach coupled with hierarchical Jaccard distance metrics. By integrating HTML tag content augmentation with unsupervised clustering algorithms, the method enables precise extraction and visual analysis of webpage structural features. This approach effectively uncovers latent similarities among phishing sites, significantly enhancing the detection of emerging and zero-day templates. Furthermore, it provides robust technical support for coordinated threat analysis, thereby bridging critical gaps in traditional defense systems regarding structured correlation analysis. Collectively, this work advances phishing detection capabilities through rigorous structural feature engineering and similarity measurement.
0 citationsRead paper
A Machine Learning Framework for Real-Time Personalized Ergonomic Pose Analysis
Traditional single-view cameras struggle to deliver comprehensive, real-time ergonomic posture assessment due to fixed viewpoints and occlusion. This work proposes a real-time analysis system that fuses multi-view 3D point clouds with 2D pose estimation, uniquely leveraging user-annotated personalized samples to train a deep learning classifier and enabling continuous inference of ergonomic postures on streaming data. By transcending the limitations of a single viewpoint, the method effectively handles occluded scenarios and achieves high-accuracy skeletal labeling and posture recognition in load-carrying tasks, demonstrating its practicality and scalability for workplace health and safety monitoring.
0 citationsRead paper
Attention-based multiple instance learning for predominant growth pattern prediction in lung adenocarcinoma wsi using foundation models
This study addresses the challenge of lung adenocarcinoma grading, which relies on accurate identification of the predominant growth pattern but typically demands costly, slide-level fine-grained annotations. To mitigate this requirement, the authors propose an attention-based multiple instance learning (MIL) framework that operates solely with whole-slide image-level labels. The approach leverages a pretrained foundation model in computational pathology—such as Prov-GigaPath—as a patch encoder and employs an attention mechanism to aggregate global features for growth pattern prediction. This work represents the first integration of a pathology foundation model with attention-based MIL, substantially reducing dependence on pixel-level annotations while enhancing prediction robustness. Experimental results demonstrate that fine-tuned Prov-GigaPath within the ABMIL framework achieves state-of-the-art performance (Cohen’s κ = 0.699), significantly outperforming conventional aggregation baselines.
0 citationsRead paper
Identification and Anonymization of Named Entities in Unstructured Information Sources for Use in Social Engineering Detection
This study addresses the challenge of balancing data availability for cybercrime analysis with privacy protection under regulations such as the GDPR by proposing an end-to-end multimodal data processing pipeline. The pipeline integrates speech enhancement, high-precision named entity recognition (NER), and structure-preserving anonymization techniques to construct a compliant dataset from Telegram-collected text, audio, and images. Experimental results demonstrate that the Parakeet model achieves optimal speech transcription performance, while the proposed Transformer-based NER approach attains the highest F1 score in identifying sensitive information. Furthermore, the anonymized data retains essential semantic structures while satisfying regulatory compliance requirements, thereby effectively supporting research on social engineering attack detection.
0 citationsRead paper