检索结果-内蒙古大学图书馆

IEEE Conference on Wireless Communications and Networking

作者： Mohannad Abu Issa Mohamed Ibnkahla Ashraf Matrawy Abdelrahman Eldosouky Systems and Computer Engineering Department Carleton University Ottawa Canada

ISBN: (数字)9798350303582

ISBN: (纸本)9798350303599

Machine learning-based intrusion detection systems (IDSs) serve as a defense-in-depth layer for Internet of Things (IoT) networks by detecting potential intrusions within IoT traffic. However, resource-constrained IoT devices impose sig-nificant challenges in developing effective IDSs. Recently, fed-erated learning (FL) has emerged as a promising solution for training detection models on distributed IoT devices without compromising resource limitations resulting in the introduction of FL-based IDSs. To this end, this paper introduces a novel approach to enhance the effectiveness of current FL-based IoT IDSs while utilizing the same resources. The proposed system partitions the FL rounds between IoT device groups, allowing each group to update the FL detection model during its time partition. Hence, by implementing this temporal partitioning, multiple detection models are updated within one FL round using the same IoT resources. The main design goals of the proposed approach are to improve detection accuracy and convergence time compared to the traditional approach. For this purpose, the proposed approach is evaluated and compared to the traditional approach using five intrusion scenarios on IoT traffic obtained from the Edge-IIoTset dataset. The results demonstrate that the proposed temporal partitioned FL-based IoT IDS outperforms the traditional system by achieving higher detection accuracy and faster convergence time. Furthermore, the proposed approach achieves the required detection accuracy in fewer FL rounds, which can, in principle, save more IoT resources.

关键词： Training Accuracy Federated learning Image edge detection Intrusion detection Internet of Things Convergence

来源：评论

学校读者我要写书评

暂无评论

Fast NMPC Design for Image-Based Visual Servoing of Autonomous Underwater Vehicles

Fast NMPC Design for Image-Based Visual Servoing of Autonomo...

引用

Industrial Cyber-Physical systems (ICPS)

作者： Hang Gu Chao Shen Department of Systems and Computer Engineering Carleton University Ottawa Canada

ISBN: (数字)9798350363012

ISBN: (纸本)9798350363029

This paper presents a new approach to image-based visual servoing (IBVS) for Autonomous Underwater Vehicles (AUVs) with the goal of improved performance and computational efficiency. Traditional IBVS methods, when combined with Model Predictive Control (MPC), face high computational demands due to the nonlinear dynamics and the large degrees of freedom (DOFs) in the variables of the associated optimization problem. Our method addresses this by reducing the DOFs of the optimization variable in the cost function while maintaining a good control performance. To further consider the smoothness of the MPC control signal, a soft constraint handling method is developed. The fast nonlinear MPC, combined with smoother control trajectories and effective constraint handling, makes our method particularly suitable for AUV IBVS applications in dynamic environments. Comparisons with standard strategies confirm the improved performance of our approach in terms of both speed and trajectory quality. Simulation results show that our approach can achieve an improved computation up to 100 times faster than conventional MPC-based IBVS methods, which highlights the great potential for real-time IBVS applications.

关键词： Constraint handling Autonomous underwater vehicles Service robots Simulation Real-time systems Visual servoing Computational efficiency

来源：评论

学校读者我要写书评

暂无评论

LookHere: Vision Transformers with Directed Attention Generalize and Extrapolate

arXiv

引用

arXiv 2024年

作者： Fuller, Anthony Kyrollos, Daniel G. Yassin, Yousef Green, James R. Department of Systems and Computer Engineering Carleton University OttawaON Canada

High-resolution images offer more information about scenes that can improve model accuracy. However, the dominant model architecture in computer vision, the vision transformer (ViT), cannot effectively leverage larger images without finetuning — ViTs poorly extrapolate to more patches at test time, although transformers offer sequence length flexibility. We attribute this shortcoming to the current patch position encoding methods, which create a distribution shift when extrapolating. We propose a drop-in replacement for the position encoding of plain ViTs that restricts attention heads to fixed fields of view, pointed in different directions, using 2D attention masks. Our novel method, called LookHere, provides translation-equivariance, ensures attention head diversity, and limits the distribution shift that attention heads face when extrapolating. We demonstrate that LookHere improves performance on classification (avg.↑ 1.6%), against adversarial attack (avg.↑ 5.4%), and decreases calibration error (avg.↓ 1.5%) — on ImageNet without extrapolation. With extrapolation, LookHere outperforms the current SoTA position encoding method, 2D-RoPE, by 21.7% on ImageNet when trained at 2242 px and tested at 10242 px. Additionally, we release a high-resolution test set to improve the evaluation of high-resolution image classifiers, called ImageNet-HR. © 2024, CC BY.

关键词： Distribution transformers

来源：评论

学校读者我要写书评

暂无评论

Pedestrian Detection: An Explainable Approach

Pedestrian Detection: An Explainable Approach

引用

International Conference on Machine Learning and Applications (ICMLA)

作者： Hongbo Pang Changcheng Huang Department of Systems and Computer Engineering Carleton University Ottawa Canada

ISBN: (数字)9798350374889

ISBN: (纸本)9798350374896

In the rapidly advancing field of autonomous technologies, particularly in transportation and surveillance, there is a critical need for accurate and reliable pedestrian detection systems. While deep learning models have achieved remarkable success in this area, their “black box” nature often concerns their trustworthiness and practical applicability. Our study introduces a modified YOLOv5s model, augmented with a Global Attention Mechanism and enhanced with Gradient-weighted Class Activation Mapping (Grad-CAM) for improved explainability. This approach maintains the high accuracy typical of deep learning models while significantly increasing their transparency. Our modified YOLOv5s architecture is tailored for real-time applications, achieving an inference speed of 9.4 milliseconds on a single NVIDIA A100 GPU. Incorporating Grad-CAM provides a more intuitive interpretation of the model's decisions, potentially increasing user confidence in its predictions. We evaluate our enhanced framework on the CityPersons dataset, demonstrating improvements in precision, recall, and mAP@0.5 compared to the baseline YOLOv5s model. Our results indicate a promising advancement in pedestrian detection, balancing high performance with improved explainability.

关键词： YOLO Deep learning Pedestrians Accuracy Surveillance Transportation Graphics processing units Predictive models Real-time systems Reliability

来源：评论

学校读者我要写书评

暂无评论

A Blockchain-based Dual Identity Management and Authentication Framework for IoT Networks

A Blockchain-based Dual Identity Management and Authenticati...

引用

IEEE World Forum on Internet of Things (WF-IoT)

作者： Dana Haj Hussein Ethan Houlahan Alexandre Janelle-Goode Thomas Lumsden Mohamed Ibnkahla Department of Systems and Computer Engineering Carleton University Ottawa Canada

ISBN: (数字)9798350373011

ISBN: (纸本)9798350373028

The widespread adoption of Internet of Things (IoT) devices has introduced an era of unparalleled connectivity and data-driven innovation. Nevertheless, this rapid proliferation introduced significant security challenges. The limited computational capabilities of IoT devices impose constraints on the implementation of robust security mechanisms, rendering them susceptible to malicious attacks. Additionally, the dynamic nature of IoT systems which allows multiple administrators to add or remove IoT devices, necessitates the establishment of an identity management system. This system is crucial for managing both devices and system users, ensuring the traceability of device registration activities, and maintaining accountability of system users. Furthermore, traditional centralized authentication systems encounter scalability issues and high costs associated with centralized servers. To address these challenges, this paper proposes a Dual Identity Management and Authentication (DIMA) framework, leveraging blockchain technology. The proposed framework addresses two primary issues. Firstly, it assigns dual identities to each IoT node, facilitating a secure authentication process and regulating network access. Secondly, it supports the traceability of system device registration activities, enabling accountability for system users and ensuring a secure administrator-in-the-loop device identification process. Moreover, experimental testing and security analysis are undertaken to validate the usability and ascertain the critical security strengths of the DIMA framework.

关键词： Technological innovation Authentication Throughput User experience Internet of Things Security Time factors Servers Usability Testing

来源：评论

学校读者我要写书评

暂无评论

COVID-19 Classification Using Pre-Trained Models and Disease Severity Score Masks

COVID-19 Classification Using Pre-Trained Models and Disease...

引用

Computing, Internet of Things and Microwave systems (ICCIMS), International Conference on

作者： Ebrahim A. Nehary Sreeraman Rajan Carlos Rossa Department of Systems and Computer Engineering Carleton University Ottawa Canada

ISBN: (数字)9798350351736

ISBN: (纸本)9798350351743

The significance of early detection of COVID-19 has been widely acknowledged as a means of reducing its spread and mortality rates among patients. Deep learning techniques for COVID-19 classification based on ultrasound (US) data have been extensively employed. However, detecting COVID-19 based on US images continues to be challenging primarily due to limited datasets with noisy and low-resolution images. This study investigates methods to enhance classification performance by incorporating disease severity score masks while training pretrained models enhanced with self-attention mechanisms. The disease severity scores range from 0 for healthy lung tissue to 1 for initial signs of abnormality, and 2 and 3 for advanced pathological artifacts. These masks and their corresponding US images are employed as inputs to pre-trained models for feature extraction. Subsequently, features extracted from the masks are utilized to recalibrate features obtained from US images using self-attention mechanisms. The proposed method achieves classification accuracy of 95.4, 90.4, 95%, 83%, and 92% when using pre-trained models VGG16, NASNet-Mobile, MobileNet_V2, ResNet50, and Xception, respectively. Further, all pre-trained models yield a low standard deviation of less than 5%. The results demonstrate that incorporating disease severity masks improves the classification performance, thus offering promising techniques for enhancing COVID-19 detection using ultrasound imaging.

关键词： COVID-19 Training Pathology Ultrasonic imaging Accuracy Computational modeling Feature extraction Standards Diseases Residual neural networks

来源：评论

学校读者我要写书评

暂无评论

Comparative Analysis of Real-Time Data Processing Architectures: Kafka versus MQTT Broker in IoT

Comparative Analysis of Real-Time Data Processing Architectu...

引用

Electronic Communications, Internet of Things and Big Data (ICEIB), International Conference on

作者： Chung Lai Deryck Ho Chung–Horng Lung Zhengding Mao Department of Systems and Computer Engineering Carleton University Ottawa Canada

ISBN: (数字)9798350360721

ISBN: (纸本)9798350360738

The exponential growth of Internet of Things (IoT) devices highlights the critical need for robust and efficient messaging systems to ensure seamless communication and real-time data exchange. In this research, a thorough comparative analysis was conducted focusing on the real-time data processing architectures of two leading messaging brokers in the IoT domain: Apache Kafka and Message Queuing Telemetry Transport (MQTT) broker. Despite widespread adoption, Kafka and MQTT brokers exhibit uniqueness in their architectures, features, and functionality in real-time data processing scenarios. The study aims to unravel the strengths and weaknesses of each technology within the context of real-time data processing, providing valuable insights to optimize their IoT data workflows.

关键词： Focusing Big Data Data processing Real-time systems Internet of Things Telemetry Queueing analysis

来源：评论

学校读者我要写书评

暂无评论

Sensor Fusion DEVS for Angle Estimation on Inertial Measurement Unit 23

Sensor Fusion DEVS for Angle Estimation on Inertial Measurem...

引用

Proceedings of the Winter Simulation Conference

作者： Gabriel Wainer Joseph Boi-Ukeme Vedant Paranjape Department of Systems and Computer Engineering Carleton University Ottawa Canada

ISBN: (纸本)9798350369663

We explore the application of a Sensor Fusion Framework, called SAFE (Simple, Applicable, Extensible, and Flexible) to improve the reliability of measurements obtained from Inertial Measurement Unit (IMU) sensors. SAFE is built using a DEVS specification and the Cadmium tool. Measuring angular position is a difficult task due to the unreliability of gyroscopes and accelerometers, two sensors widely used to measure angles. Although angular position can be measured using imaging systems, these are costly, and not ideal for handheld and portable devices. An alternative solution is to use sensor fusion to fuse the readings of both accelerometer and gyroscope, obtaining reliable readings. We show the application of the SAFE methodology and the results of our case study showing the potential of this method.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Cross-Layer Management Framework for Enhancing XR-Based System Security in Zero-Trust Wireless Communications

引用

IEEE Journal on Selected Areas in Communications 2025年第6期43卷 2011-2024页

作者： Ghourab, Esraa M. Azab, Mohamed Gracanin, Denis Al-Qutayri, Mahmoud Muhaidat, Sami Khalifa University 6G Center Department of Computer Science Abu Dhabi127788 United Arab Emirates Virginia Military Institute Department of Computer and Information Sciences VA United States Virginia Tech Department of Computer Science BlacksburgVA24060 United States Center Department of Computer and Information Engineering Abu Dhabi127788 United Arab Emirates Khalifa University 6G Center Department of Computer and Information Engineering Abu Dhabi127788 United Arab Emirates Carleton University Department of Systems and Computer Engineering OttawaONK1S 5B6 Canada

Extended reality (XR) and 6G networks are set to transform mobile immersive experiences, with privacy and security being paramount in XR communications. Achieving secure and reliable XR experiences while meeting high-resolution and low-latency requirements is challenging for wireless networks. A novel security-aware cross-layer communication management framework is proposed, employing zero-trust spatiotemporal physical layer level manipulations for moving-target defense. Driven by deep reinforcement learning and real-time monitoring, the proposed framework adaptively reprograms the network configuration to maximize the user’s quality of experience (QoE), reduce the overall latency, and minimize the attacker’s intercept probability. The framework was evaluated in a simulated scenario featuring an indirect multi-hop communication setup. The results show that the proposed framework effectively and efficiently secures XR user communications while maintaining QoE, outperforming conventional Q-learning algorithms. © 1983-2012 IEEE.

关键词： Resource allocation

来源：评论

学校读者我要写书评

暂无评论

A Game-Theoretic Approach for Security Control Selection

arXiv

引用

arXiv 2024年

作者： Léveillé, Dylan Jaskolka, Jason Department of Systems and Computer Engineering Carleton University OttawaON Canada

Selecting the combination of security controls that will most effectively protect a system’s assets is a difficult task. If the wrong controls are selected, the system may be left vulnerable to cyberattacks that can impact the confidentiality, integrity and availability of critical data and services. In practical settings, it is not possible to select and implement every control possible. Instead considerations, such as budget, effectiveness, and dependencies among various controls, must be considered to choose a combination of security controls that best achieve a set of system security objectives. In this paper, we propose a game-theoretic approach for selecting effective combinations of security controls based on expected attacker profiles and a set budget. The control selection problem is set up as a two-person zero-sum one-shot game. Valid control combinations for selection are generated using an algebraic formalism to account for dependencies among selected controls. We demonstrate the proposed approach on an illustrative financial system used in government departments under four different scenarios. The results illustrate how a security analyst can use the proposed approach to guide and support decision-making in the control selection activity when developing secure systems. © 2024, CC BY.

关键词： Budget control

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：