記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

This study offers a controlled comparison of facerecognition pipelines on Indonesian Muslim Face Dataset under a single, fixed protocol designed to minimize confounds. Faces are detected and aligned with Multi-task Cascaded Convolutional Network, data are split once using a stratified 80:20 partition. The training split undergoes class balancing and moderate occlusion exposure, while the test split remains untouched and is evaluated with single-crop inference. We examine two representation families: identity-preserving embeddings from ArcFace (ResNet-100, buffalo_l) and frozen global descriptors from EfficientNet-B0 and DenseNet-121. Each representation is paired with four classical classifiers, Support Vector Classifier, K-Nearest Neighbors, Random Forest, and Decision Tree, and assessed using macro-averaged accuracy, precision, recall, and F1 score. Results show that ArcFace produces compact, well-separated identity clusters, yielding near ceiling performance across classifiers, whereas frozen global descriptors are weaker and markedly more sensitive to the choice of decision rule (ensembles tend to suit EfficientNet-B0, while a margin-based SVC aligns better with DenseNet-121). Confusion matrix analyses corroborate these trends, revealing sparse residual confusions for ArcFace and broader, asymmetric error patterns for the global descriptors, especially under partial occlusion. The study offers a reproducible baseline and practical guidance for small-tomedium, imbalanced settings: identity-preserving embeddings coupled with a simple linear margin or KNN provide a dependable default, where only generic CNNs are available, tree ensembles and margin-based methods should be chosen to match descriptor geometry.

DOI： 10.1109/IMCOM69009.2026.11360880

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=105033801088&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Recently, growing public concern over data privacy has led to the emergence of the ‘right to be forgotten’ under the European Union’s General Data Protection Regulation (GDPR), which allows individuals to request the deletion of personal data. In response, researchers have developed methods to selectively remove or mitigate the influence of specific data on machine learning models, a process known as machine unlearning (MUL). However, while most research on MUL has focused exclusively on maintaining model accuracy after unlearning, the robustness of these models has been largely overlooked, especially under adversarial attacks. In this study, we examine whether unlearning degrades robustness in the visual classification task by performing adversarial attacks on various unlearned methods and compare performance against a retrained-from-scratch baseline, including evaluations using a standard corruption dataset. Our comprehensive evaluation across seven unlearning methods reveals consistent degradation in robustness under perturbation, supporting our hypothesis that machine unlearning degrades adversarial robustness compared to retraining from scratch, and highlighting the need for methods that preserve robustness without compromising the effectiveness and efficiency.

DOI： 10.1007/978-981-95-7820-7_15

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=105039947587&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

We present the eyeglass-type switch, an eyeglass-mounted eye/blink switch designed for nurse-call operation by people with severe motor impairments, with a particular focus on amyotrophic lateral sclerosis (ALS). The system targets real-world bedside constraints—low illumination at night, supine posture, and network-independent operation—by combining near-infrared (NIR) LED illumination with an NIR eye camera and executing all processing on a small, GPU-free computer. A two-stage convolutional pipeline estimates eight periocular landmarks and the pupil center; eye-closure is detected either by a binary classifier or by an angle criterion derived from landmarks, which also skips pupil estimation during closure. User intent is determined by crossing a caregiver-tunable “off-area” around neutral gaze, implemented as rectangular or sector shapes. Four output modes—single, continuous, long-press, and hold-to-activate—are supported for both oculomotor and eyelid inputs. Safety is addressed via relay-based electrical isolation from the nurse-call circuit and audio feedback for state indication. The prototype runs at 18 fps on commodity hardware. In feature-point evaluation, mean errors were 2.84 pixels for landmarks and 1.33 pixels for the pupil center. In a bedside task with 12 healthy participants, the system achieved (Formula presented.) in single mode and (Formula presented.) in hold-to-activate mode; blink-only input yielded (Formula presented.). Performance was uniformly high for right/left/up and eye-closure cues, with lower recall for downward gaze due to eyelid occlusion, suggesting camera placement or threshold tuning as remedies. The results indicate that the proposed switch provides reliable, low-burden nurse-call control under nighttime conditions and offers a practical input option for emergency alerts and augmentative and alternative communication (AAC) workflows.

DOI： 10.3390/electronics14214201

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=105021396686&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

This study presents a comparative evaluation of deep feature extraction and machine learning classifiers for facial recognition-based student attendance systems. Two pretrained convolutional neural networks, EfficientNetB0 and DenseNet121, were used to extract visual features from facial images collected from 21 individuals. The extracted features were classified using Support Vector Classifier (SVC), KNearest Neighbors (KNN), Random Forest, and Decision Tree. Prior to classification, the dataset was balanced through oversampling and undersampling techniques and further enhanced with occlusion-based augmentation to improve robustness. The results show that the combination of EfficientNetB0 and SVC consistently outperformed other configurations, achieving an accuracy of 86.68 percent and demonstrating better generalization across all classes. Compared to DenseNet121, EfficientNetB0 produced more compact and discriminative features, particularly effective when paired with margin-based classifiers such as SVC. Traditional models such as Decision Tree yielded significantly lower performance, especially on high-dimensional features. Additionally, the propose approach surpassed several baseline methods in previous studies, offering a competitive and computationally efficient alternative to end-to-end deep learning systems. These findings emphasize the effectiveness of combining lightweight deep feature extraction with classifiers, particularly for real-time and resource-constrained facial recognition applications.

DOI： 10.1109/ICEEIE66203.2025.11251838

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=105029908825&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Gaze attention estimation is the task that aims to understand where each person is looking in each scene. In this study, we introduce a new annotated dataset that is derived from medical simulation training videos, capturing diverse and authentic clinical scenarios from a practical medical environment and annotated by the ground truth data from eye-tracking devices (iMotions [1]) worn by medical professionals during procedures in the scenes. Most of the existing approaches in the field of gaze prediction often rely on object detection as a guide for the model. However, this becomes problematic when encountering specialised tools and equipment in medical environments, and that domain is probably absent from the standard large-scale object detection and segmentation dataset. To address the problem, this paper attempts to propose a gaze prediction framework that integrates with the head pose information, which consists of pitch, yaw, and roll. Enable the model to rely on gaze direction even when objects are not detected. This approach is based on the self-attention mechanism of vision transformers. We believe this will enhance the model's performance in relation to the relationship between gaze direction and the scene. We hope to offer a more reliable framework for real-world medical applications.

DOI： 10.23919/MVA65244.2025.11175121

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=105019057700&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

Micro-expressions are very brief, involuntary facial expressions that reveal hidden emotions, lasting less than a second, while macro-expressions are more prolonged facial expressions that align with a person’s conscious emotions, typically lasting several seconds. Micro-expressions are difficult to detect in lengthy videos because they have tiny amplitudes, short durations, and frequently coexist alongside macro-expressions. Nevertheless, micro- and macro-expression analysis has sparked interest in researchers. Existing methods use optical flow features to capture the temporal differences. However, these optical flow features are limited to two successive images only. To address this limitation, this paper proposes LGNMNet-RF, which integrates a Lite General Network with MagFace CNN and a Random Forest classifier to predict micro-expression intervals. Our approach leverages Motion History Images (MHI) to capture temporal patterns across multiple frames, offering a more comprehensive representation of facial dynamics than optical flow-based methods, which are restricted to two successive frames. The novelty of our approach lies in the combination of MHI with MagFace CNN, which improves the discriminative power of facial micro-expression detection, and the use of a Random Forest classifier to enhance interval prediction accuracy. The evaluation results show that this method outperforms baseline techniques, achieving micro-expression F1-scores of 0.3019 on CAS(ME)² and 0.3604 on SAMM-LV. The results of our experiment indicate that MHI offers a viable alternative to optical flow-based methods for micro-expression detection.

DOI： 10.3390/a17110491

Kyutacar

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85210317376&origin=inward

担当区分：筆頭著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Adversarial attacks pose a significant threat to the reliability and security of deep learning models, particularly in image processing applications. Defending against these sophisticated manipulations requires innovative strategies, with Generative Adversarial Networks (GANs) emerging as a promising solution. This paper presents an experimental exploration of the power of conditional Generative Adversarial Networks (cGANs) in image reconstruction-based strategies for defending against adversarial attacks. Our study involves a comparative analysis of four distinct image reconstruction models: the traditional GAN-based Defense-GAN, the cGAN-based method exemplified by pix2pix, a hybrid approach combining pix2pix with perceptual loss, and a generator model centered around residual blocks. The results of our experiments demonstrate that cGAN models exhibit significantly enhanced efficacy in defending against adversarial attacks compared to other image reconstruction methods. This superiority is attributed to the inherent characteristics of cGANs, which we delve into in detail. The findings provide crucial insights for developing more robust defense strategies against adversarial attacks in diverse image processing and machine learning applications.

DOI： 10.1007/978-3-031-57870-0_14

DOI： 10.1007/978-3-031-57870-0_14

Kyutacar

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

It is imperative to think about the security requirements of humanizing artificial intelligence because facial expression recognition (FER) is necessary for human-computer interaction. According to past studies, adding relatively modest perturbations to the input vector makes it simple to change the output of deep learning (DL)-based models. However, research on adversarial examples that target FER systems is still in its infancy. Thus, we analyze a black-box attack under a very constrained condition, where only one pixel can be changed in this study. To do this, we suggest a novel technique based on particle swarm optimization (PSO) that generates adversarial perturbations at the level of a single pixel in a superpixel of a face image to fool two popular DL-based FER systems, such as FER-net, ResNet50, and VGG16. All the experiments are performed on three publicly available benchmark datasets, such as Japanese Female Facial Expression (JAFFE), Extended Cohn-Kanade (CK+), and FED-RO, in two various attack scenarios, such as untargeted and targeted. On the one hand, the success rates of the proposed method on JAFFE, CK+, and FED-RO are 62.5%, 26.12%, and 42.5%, respectively, while fooling FER-net in untargeted attack scenarios. On the other hand, the success rates of the proposed method on JAFFE, CK+, and FED-RO are 77.14%, 50%, and 48.71%, respectively, while fooling VGG16 in an untargeted attack scenario. The findings demonstrate that the proposed method outranks one well-known differential evolution-based pioneering approach.

DOI： 10.1109/DSC63325.2024.00016

Kyutacar

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85214561554&origin=inward

担当区分：筆頭著者,　責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

Recently, an increasing number of laws have governed the useability of users’ privacy. For example, Article 17 of the General Data Protection Regulation (GDPR), the right to be forgotten, requires machine learning applications to remove a portion of data from a dataset and retrain it if the user makes such a request. Furthermore, from the security perspective, training data for machine learning models, i.e., data that may contain user privacy, should be effectively protected, including appropriate erasure. Therefore, researchers propose various privacy-preserving methods to deal with such issues as machine unlearning. This paper provides an in-depth review of the security and privacy concerns in machine learning models. First, we present how machine learning can use users’ private data in daily life and the role that the GDPR plays in this problem. Then, we introduce the concept of machine unlearning by describing the security threats in machine learning models and how to protect users’ privacy from being violated using machine learning platforms. As the core content of the paper, we introduce and analyze current machine unlearning approaches and several representative results and discuss them in the context of the data lineage. Furthermore, we also discuss the future research challenges in this field.

DOI： 10.1007/s42979-023-01767-4

Kyutacar

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85153474758&origin=inward

担当区分：筆頭著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Convolutional neural networks are widely used for image recognition tasks, but they are vulnerable to adversarial attacks that can cause the model to misclassify an image. Such attacks pose a significant security risk in safety-critical applications like facial recognition and autonomous driving. Researchers have made progress in defending against adversarial attacks through two approaches: enhancing the neural networks themselves to be more robust and removing the perturbation added to the image through pre-processing. This paper is based upon a recent defense model that belongs to the latter approach, which utilizes image-to-image translation to regenerate images perturbed by adversarial attacks. We optimized the training process of their model and tested the model performance against more recent and strong attacks. The results show that the model is able to regenerate images attacked by the state-of-the-art attack, the AutoAttack, and restores the classification accuracy to a level over 83% to that of the original images.

DOI： 10.1007/978-3-031-41181-6_32

Kyutacar

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85174448356&origin=inward

担当区分：筆頭著者,　責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

Deep learning has become one of the most popular research topics today. Researchers have developed cutting-edge learning algorithms and frameworks around deep learning, applying them to a wide range of fields to solve real-world problems. However, we are more concerned about the security risks associated with deep learning models, such as adversarial attacks, which this article will discuss. Attackers can use the deep learning model to create the conditions for an attack, maliciously manipulating the input images to deceive the classification model and produce false positives. This paper proposes a method of pre-denoising all input images to prevent adversarial attacks by adding a purification layer before the classification model. The method in this paper is proposed based on the basic architecture of Conditional Generative Adversarial Networks. It adds the image perception loss to the original algorithm Pix2pix to achieve more efficient image recovery. Our method can recover noise-attacked images to a level close to the actual image to ensure the correctness of the classification results. Experimental results show that our approach can quickly recover noisy images, and the recovery accuracy is 20.22% higher than the previous state-of-the-art.

DOI： 10.1109/ACCESS.2021.3137637

Kyutacar

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85122080076&origin=inward

担当区分：筆頭著者,　責任著者   記述言語：英語   掲載種別：記事・総説・解説・論説等（学術雑誌）

For software engineering, if two code fragments are closely similar with minor modifications or even identical due to a copy-paste behavior, that is called software/code clone. Code clones can cause trouble in software maintenance and debugging process because identifying all copied compromised code fragments in other locations is time-consuming. Researchers have been working on code clone detection issues for a long time, and the discussion mainly focuses on software engineering management and system maintenance. Another considerable issue is that code cloning provides an easy way to attackers for malicious code injection. A thorough survey work of code clone identification/detection from the security perspective is indispensable for providing a comprehensive review of existing related works and proposing future potential research directions. This paper can satisfy above requirements. We review and introduce existing security-related works following three different classifications and various comparison criteria. We then discuss three further research directions, (i) deep learning-based code clone vulnerability detection, (ii) vulnerable code clone detection for 5G-Internet of Things devices, and (iii) real-time detection methods for more efficiently detecting clone attacks. These methods are more advanced and adaptive to technological development than current technologies, and still have enough research space for future studies.

DOI： 10.1109/ACCESS.2021.3065872

Kyutacar

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85103168898&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Japan   Kyoto   2025年07月26日  -  2025年07月28日

Kyutacar

記述言語：英語   掲載種別：研究論文（学術雑誌）

Advanced Persistent Threats (APTs) involve attackers maintaining a long-term presence on victim systems, leading to the stealthy exfiltration of sensitive data during network transfers. Despite existing methods to detect and halt APT data exfiltration, these attacks continue to pose significant threats to sensitive information and result in substantial commercial losses. Current approaches primarily focus on preemptive measures, which are insufficient once early-stage detection fails due to a lack of continuous monitoring. We propose an effective and efficient network monitoring method to address this gap and detect APT exfiltration during data transfer. Our approach assumes the presence of an undetected APT attacker within the victim system. We examine data exfiltration across three exfiltration traffic environments: exfiltration over command control channels, exfiltration over transfer size limitations, and their combinations. We introduce two detection metrics: Package Transfer Rate and Byte Transfer Rate. Utilizing these metrics, we measure network traffic, categorize APT attack environments, and train deep neural network models, named EDXGB, using ensembled decision trees to predict APT exfiltration. Our method is validated on two public datasets and compared against six baseline methods. Additionally, we simulate real-world exfiltration scenarios by creating three exfiltration traffic environments for each dataset. The results demonstrate that our method effectively detects APT exfiltration across various network environments, enhancing data protection and secure transfer. The code is open source and available at https://github.com/cxjuan/EDXGB-for-APT.

DOI： 10.1109/ACCESS.2025.3567772

Kyutacar

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=105004683219&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

This paper presents a method for analyzing Japanese Diet deliberations video footage to determine if a speaker reads from a script while speaking. The study focuses on classifying script reading faces as a preprocessing step for future research. Four main classification methods are proposed: (1) using the success or failure of face detection, (2) utilizing head pose angle based on the PnP problem, (3) employing head pose angle based on a regression model, and (4) using an image recognition model. Twelve videos from a Video Retrieval System for Diet Deliberations were used, and 30-s speaking scenes were extracted to evaluate the proposed approaches. The results indicate that approach (4) achieved the highest classification accuracy.

DOI： 10.1007/978-981-96-1535-3_20

DOI： 10.1007/978-981-96-1535-3_20

Kyutacar

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85219209279&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Japan   Kitakyushu   2024年11月14日  -  2024年11月15日

Kyutacar

担当区分：筆頭著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Japan   Kitakyushu   2024年11月14日  -  2024年11月15日

Kyutacar

担当区分：筆頭著者   記述言語：日本語   掲載種別：記事・総説・解説・論説等（学術雑誌）

DOI： 10.11517/jjsai.38.2_197

DOI： 10.11517/jjsai.38.2_197

Kyutacar

担当区分：筆頭著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Over the past decade, the development of both adversarial attack methods and defense strategies has accelerated rapidly. Classification accuracy has been predominantly used as the sole metric for assessing model performance. However, when the reported accuracy rates of two models are identical or very similar, it becomes challenging to determine which model is superior. To address this issue and offer more insights into model performance, this study introduces a novel classification performance metric: the confidence gap. This metric is defined as the difference in confidence level between the true label and either the top 1 prediction or the second-best prediction, depending on the accuracy of the image classification. The confidence level, as indicated by its sign, reflects the correctness of the classification and provides more detailed information on the robustness of the classification result. Recognizing that evaluation results may be inconsistent when employing different criteria, we recommend that future research in this field should report the confidence gap alongside accuracy rates.

DOI： 10.1007/978-3-031-41181-6_41

Kyutacar

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85174449307&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Temporal Convolution Network(TCN) has recently been introduced in the cybersecurity field, where two types of TCNs that consider causal relationships are used: causal TCN and non-causal TCN. Previous researchers have utilized causal and non-causal TCNs separately. Causal TCN can predict real-time outcomes, but it ignores traffic data from the time when the detection is activated. Non-causal TCNs can forecast results more globally, but they are less real-time. Employing either causal TCN or non-causal TCN individually has its drawbacks, and overcoming these shortcomings has become an important topic. In this research, we propose a method that combines causal and non-causal TCN in a contingent form to improve detection accuracy, maintain real-time performance, and prevent long detection time. Additionally, we use two datasets to evaluate the performance of the proposed method: NSL-KDD, a well-known dataset for evaluating network intrusion detection systems, and MQTT-IoT-2020, which simulates the MQTT protocol, a standard protocol for IoT machine-to-machine communication. The proposed method in this research increased the detection time by about 0.1ms compared to non-causal TCN when using NSL-KDD, but the accuracy improved by about 1.5%, and the recall improved by about 4%. For MQTT-IoT-2020, the accuracy improved by about 3%, and the recall improved by about 7% compared to causal TCN, but the accuracy decreased by about 1% compared to non-causal TCN. The required time was shortened by 30ms (around 30%), and the recall was improved by about 7%.

DOI： 10.1007/978-3-031-45933-7_30

Kyutacar

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85178520287&origin=inward

担当区分：筆頭著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Japan   Kitakyushu   2020年02月24日  -  2020年02月26日

This exploratory investigation aims to discuss current network environment of digital supply chain system and security issues, especially from the Internet world, of digital supply chain management system with applying some advanced information technologies, such as Internet of Things and blockchain, for improving various system performance and properties. This paper introduces the general histories and backgrounds, in terms of information science, of the supply chain and relevant technologies which have been applied or are potential to be applied on supply chain with purpose of lowering cost, facilitating its security and convenience. It provides a comprehensive review of current relative research work and industrial cases from several famous companies. It also illustrates the IoT enablement and security issues of current digital supply chain system, and existing blockchain’s role in this kind of digital system. Finally, this paper concludes several potential or existing security issues and challenges which supply chain management is facing.

DOI： 10.1007/978-3-030-39746-3_57

DOI： 10.1007/978-3-030-39746-3_57

Kyutacar

担当区分：筆頭著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

This exploratory investigation aims to discuss current status and challenges, especially in aspect of security and trust problems, of digital supply chain management system with applying some advanced information technologies, such as Internet of Things, cloud computing and blockchain, for improving various system performance and properties, i.e. transparency, visibility, accountability, traceability and reliability. This paper introduces the general histories and definitions, in terms of information science, of the supply chain and relevant technologies which have been applied or are potential to be applied on supply chain with purpose of lowering cost, facilitating its security and convenience. It provides a comprehensive review of current relative research work and industrial cases from several famous companies. It also illustrates requirements or performance of digital supply chain system, security management and trust issues. Finally, this paper concludes several potential or existing security issues and challenges which supply chain management is facing.

DOI： 10.1109/DASC/PiCom/CBDCom/CyberSciTech.2019.00069

Kyutacar

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85075142259&origin=inward

開催期間： 2025年03月01日   記述言語：日本語   開催地：那覇商工会議所＆オンライン（ハイブリッド開催）  

開催期間： 2025年03月01日   記述言語：日本語   開催地：那覇商工会議所＆オンライン（ハイブリッド開催）  

開催期間： 2024年12月10日 - 2024年12月11日   記述言語：日本語   開催地：伊香保温泉ホテル天坊  

開催期間： 2024年12月10日 - 2024年12月11日   記述言語：日本語   開催地：伊香保温泉ホテル天坊  

開催期間： 2024年12月10日 - 2024年12月11日   記述言語：英語   開催地：伊香保温泉ホテル天坊  

開催期間： 2024年12月04日 - 2024年12月05日   記述言語：日本語   開催地：産業技術総合研究所 臨海副都心センター