The integration of edge computing with autonomous vehicles (AVs) has emerged as a pivotal technological advancement to address the computational, latency, and data privacy challenges inherent in autonomous navigation. This review examines the critical need for data offloading in AV systems, providing an in-depth analysis of different data offloading strategies, including full, partial, and adaptive offloading. We explore various types of edge computing nodes, ranging from onboard embedded systems to roadside and cloudlet-based architectures, highlighting their roles and trade-offs in AV ecosystems. Furthermore, the paper compares multiple Radio Access Technologies (RATs), such as 5G, mm Wave, Wi-Fi, and DSRC, emphasizing their effectiveness in supporting low-latency and high-bandwidth data offloading for various applications. The ongoing innovations in docker and containerization that has revolutionized edge computing for AVs by enabling lightweight, portable, and scalable deployment of services, reducing dependency on specific hardware platforms and improving the overall flexibility and efficiency of the system have also been discussed. A comparative analysis of different edge computing architectures is presented, focusing on centralized, decentralized, and fog computing models tailored for AVs. The review also addresses key challenges in the deployment and scaling of edge computing for AVs, including network reliability, data management, interoperability, and security concerns. Recent trends and future directions are also highlighted. This paper aims to provide a comprehensive end-to-end understanding of the role of edge computing in autonomous vehicles, offering a roadmap for future research and development in this dynamic domain.

The present work introduces a miniaturized microstrip patch antenna meticulously engineered for wireless energy transmission systems. To accomplish spurious harmonic suppression, the proposed configuration incorporates an open-circuited stub along the feed line, effectively mitigating undesired resonant modes up to the fourth harmonic order. The antenna, precisely optimized to operate at 2.45 GHz, demonstrates a 10-dB return-loss impedance bandwidth ranging from 2.3 GHz to 2.5 GHz (8.69%), with a simulated gain of 2.2 dBi and a radiation efficiency exceeding 75%. Owing to these distinguished electromagnetic attributes, the proposed antenna emerges as a potential solution for wireless power transfer (WPT) end-use systems, where compactness, spectral purity, and high efficiency are of critical importance.

Network intrusion benchmark datasets serve an essential role in improving the advancement of research in cybersecurity because they provide standardized resources for assessing the effectiveness of intrusion detection systems and associated cybersecurity solutions. This review article provides a detailed examination of the cutting-edge in network intrusion benchmark datasets, concentrating on their features, content, utilization, and implications for cybersecurity research. We systematically review a wide variety of benchmark datasets that are often utilized in the industry, which include the DARPA, KDDcup99, NSL-KDD, Kyoto, UNSW-NB15, and CICIDS-17 datasets. We analyzed each dataset, including its performance based on machine learning and deep learning models, by critically synthesizing existing literature. Additionally, we discussed the common challenges existing in intrusion detection systems. Furthermore, we provided a description of various machine learning and deep learning algorithms used for intrusion detection applications. This study aims to assist researchers in choosing suitable datasets and techniques for evaluating and benchmarking intrusion detection systems, ultimately advancing cybersecurity research and the development of reliable and efficient cybersecurity solutions.

Network intrusion detection systems are essential for defending recent computer networks from ever-evolving cyber attacks. Security is of utmost importance due to the complex and constantly changing nature of network threats. To improve the detection capabilities in network traffic, this research presents a unique method for intrusion detection by utilizing attention-based Transformer architectures. The proposed Transformer-based model offers an adaptable and reliable method for detecting sophisticated and dynamic threats by fusing the strength of the self-attention mechanism. The model is evaluated on two network intrusion benchmark datasets (NSL-KDD and UNSW-NB15). The correlation technique is used for feature extraction, and both binary and multiclass classification with and without feature extraction are performed on the datasets. The proposed model achieved over 99% accuracy, precision, and recall on the two datasets. The experimental results indicate that the proposed approach provides better results than other systems.

Intrusion detection systems are essential in providing security to network structures by detecting and reducing the possibility of cyber attacks. This work investigates the usefulness of employing a Deep neural network in combination with the oversampling technique to improve intrusion detection on the CIC-IDS2017 dataset. This dataset is a realistic benchmark dataset that consists of modern network traffic and a variety of attack types. This method uses the synthetic oversampling method to rectify the dataset’s intrinsic class imbalance, which is a significant problem in intrusion detection due to insufficient minority attack classes. Experimental findings show that combining these methods enhances detecting rates of minority attack classes while retaining excellent overall performance. This study highlights the potential for combining sophisticated resampling methods with deep learning algorithms to provide more effective intrusion detection systems to protect against a broad spectrum of cyber threats in current network configurations. The model provided better performance outcomes regarding accuracy, recall, precision, and F1-score compared to other approaches.

In today's fast-changing digital world, cyber-attacks have become more sophisticated and complex, making it challenging to accurately detect intrusions. These intrusions not only put data security at risk but also threaten the availability and integrity of important information, posing significant dangers to individuals, organizations, and society as a whole. This study proposes Machine Learning (ML) techniques for anomaly detection using the widely recognized CIC-IDS2017 dataset, which is a benchmark in cybersecurity research. The research evaluates each approach based on the performance of classifiers after preprocessing the data to remove irrelevant features that have minimal impact on attack detection. The results show that the Random Forest classifier, combined with the Correlation feature selection method, performs exceptionally well, achieving an impressive accuracy rate of 99.61%. These results show the degree of efficiency of a robust data preprocessing strategy in enhancing intrusion detection systems' capabilities in classifying attack traffic, namely DDoS, Brute Force attacks, scans, Webbased attacks, Bots, Denial of Service, and Heartbleed vulnerabilities. The comparison between total accuracy provided by these techniques makes such research valuable in terms of insights for optimizing cybersecurity defenses against dynamically correlated threats thriving in this digital age.

Intrusion detection is essential for preserving computer network security and dependability, particularly in an ever-increasing cyber-attack era. Conventional research techniques rely on specific features and machine learning approaches that can’t detect complex patterns and irregularities in network communications data. In this article, we provide a unique technique for intrusion detection which integrates ensemble feature extraction methods with Long Short Term Memory networks. To evaluate the efficacy of the suggested technique, we used benchmark network intrusion datasets (NSL-KDD and UNSW-NB15) and compared them to other available approaches. Performance indicators such as recall, F1-score, precision, and accuracy are utilized to evaluate how well the algorithm detects intrusions. The research findings indicate that ensemble feature selection using the proposed algorithm surpasses other approaches regarding accuracy and robustness. The method provided 99.56%, 98.14% of accuracy for binary classification and 98.05% and 97.47% of accuracy for multi-class classification for both NSL-KDD and UNSW-NB15 datasets. These results demonstrate the ability of ensemble-based approaches to improve the efficiency of intrusion detection devices in real-world applications.

The detection of intrusions has a significant impact on providing information security, and it is an essential technology to recognize diverse network threats effectively. This work proposes a machine learning technique to perform intrusion detection and classification using multiple feature extraction and testing using an Extreme learning machine (ELM). The model is evaluated on the two network intrusion datasets (NSL-KDD and UNSW-NB15), which consist of real-time network traffic. The arithmetic, gradient, and statistical features were extracted and evaluated with the proposed model. The method’s efficacy is assessed using accuracy, sensitivity, specificity, precision, and F1-score. The proposed method achieves 94.5%, 97.61%, 96.91%, 96.51%, and 97.05% accuracy, sensitivity, specificity, precision, and F1-score for NSL-KDD and 94.3%, 98,36%, 99.31%, 99.67% and 99.01% of accuracy, sensitivity, specificity, precision and F1-score for the UNSWNB-15 dataset respectively, which is better performance outcomes when compared to other existing works.

To meet the need for food on a worldwide scale, both in terms of quality and quantity, it is essential to protect plants generally against disease. Even while the problem with diseases is well understood, it remains difficult to quickly identify them, especially in regions where the required infrastructure is missing. One potential answer is to employ edge devices for disease diagnostics. As a proof of idea, the Raspberry Pi is utilized to highlight the potential of smaller variants on compute devices with minimal power. Smartphones are one type of gadget. Currently, 83.72% of people use cell phones worldwide. It would be advantageous to use cell phones to serve this purpose; this project proposes a framework to do so. These images depict real-world situations and are updated using a publicly available PlantVillage dataset, additional images from Google Photographs, the inaturalist website, and real-time image capturing. SqueezeNet and EfficientDet-Lite0 are two deep learning models that are employed. EfficientDet-Lite0 is utilized for object detection, such as leaves, while SqueezeNet is utilized for picture classification, such as plant and disease diagnosis. On a Raspberry Pi 4, the model that was trained can get 97.89% accuracy in less than a second during inference. This should demonstrate how cell phones and other low-powered edge devices may be used to solve this issue.

Protecting plants in general and against disease is vital to meeting the demand for food globally in both quality and quantity. Although the disease problem is vastly known, their rapid identification remains a problem, especially in parts of the world where necessary infrastructure is lacking. One solution can be using edge devices for disease diagnosis. As a proof of concept, Raspberry Pi 4 is used to showcase the capability of smaller models on low-powered and computation devices. One such device is smartphones. Already 83.72% of the world uses smartphones. Leveraging smartphones for this task would be beneficial; for that, a framework is proposed in this project. Using a public dataset, the PlantVillage dataset, added images from Google Images, inaturalist website, and clicked photos in real-time, these images are added to represent real-world scenarios. Two deep learning models are used SqueezeNet and EfficientDet-Lite0. SqueezeNet is used for image classification, i.e., plant and disease identification, and EfficientDet-Lite0 is used for object detection, i.e., leaves. The trained model can achieve 96.88% accuracy with inference time under one second on Raspberry Pi 4. This should showcase the potential of smartphones and other low-powered edge devices for this problem.

Network Intrusion Detection System has become a crucial component of the Internet of Things (IoT) framework for expanding Internet security problems. Most intrusion detection research in the past used the KDDCUP99 dataset for testing. However, certain prevalent instances still need to be included when comparing the KDDCUP99 dataset to the UNSWNB15 dataset to evaluate Network Intrusion Detection System. For classification studies, we present Random Forest, Logistic Regression, and Support Vector Machine techniques in this study. Our method's effectiveness is assessed using Precision, F1-Score, Recall, and accuracy. The findings are evaluated in light of previously conducted studies. The experimental findings demonstrate the suggested Random forest method provides better results than alternative approaches.

Network intrusion has emerged as an essential issue for business and government societies in the cyber-threat environment. Implementing a system for detecting network intrusions has been recognized as crucial to distinguish between regular and anomalous network traffic to mitigate this potential risk. The efficiency of an intrusion detection system in a network is determined by its ability to identify potential threats as anomalies appropriately. This research primarily examines Intrusion Detection Systems (IDSs) constructed utilizing machine learning techniques. This study evaluates the UNSW-NB15 network intrusion dataset, which will be utilized for training and testing the algorithms. Additionally, a filter-based feature selection strategy is employed with the XGBoost algorithm. Next, we proceed to construct the following machine-learning algorithms, making use of the decreased feature space: Artificial neural network, k-Nearest Neighbour, and Decision Tree. Both binary and multiclass classifications are evaluated, and the results are compared with previous approaches.

Using real-time data to measure blood pressure from single photoplethysmography (PPG) is proposed. Due to an increase in publicly available datasets, the application of machine learning techniques in medical research studies has expanded in recent years. The Datasets utilized in this work were taken from the Queensland Vital Signs Dataset. Five feature vectors from the Photoplethysmography (PPG) signal are extracted and sampled at a rate of 25Hz. The R2 score value and Mean Square Error (MSE) are used to measure the performance of the various models. The best results for Systolic Blood Pressure are 0.91 and 7.76, respectively. While for Diastolic Blood Pressure the best results achieved were 0.87 and 7.06, respectively.