Open Radio Access Network (O-RAN) architectures are driving a shift towards modular, software-defined RANs, where virtualization is key for flexible component development. OpenAirInterface (OAI) provide software-based RAN functionalities, complemented by Near-Real-Time RAN Intelligent Controllers (Near-RT RICs) like FlexRIC. However, the performance of these virtualized environments is fundamentally limited by the underlying compute infrastructure. This study establishes a crucial performance baseline by isolating the impact of raw computational constraints within a virtualized O-RAN testbed composed of OAI simulators and the FlexRIC controller. To achieve this, our setup deliberately excludes radio virtualization middleware, ensuring all signal processing and channel emulation to be offloaded to the CPU. We assess the impact of these constraints on the end-user’s Quality of Experience (QoE) by streaming video and audio files to a client within the simulated UE’s network namespace. Our findings reveal that video quality degrades significantly, as reflected in low VMAF, while audio quality holds steady under the same constraints.

Ransomware attacks have significantly risen recently, becoming more complex and posing a greater global threat. The contribution of machine learning in this domain is significant. Our literature review reveals that the field of ransomware severity prediction is under-explored, leading to a comparative study of predictive models for ransomware severity. We analyze over 50 verified ransomware strains using the MITRE ATT&CK framework and CAR for behavior and severity analysis and prepare a dataset. This study evaluates six machine learning models to predict ransomware attack severity, using regression and classification methods to identify the most effective predictive model. Performance tests reveal that the Neural Network is the most effective method of predicting ransomware severity, followed by Support Vector Machine (SVM) and XGBoost (XGB), underscoring the importance of machine learning in enhancing cybersecurity and proactive threat severity management. This research also reveals that Yanluowang Ransomware has the highest average severity score, indicating that it is more severe and harmful than other ransomware families. The most frequent ransomware family is identified as LokiLocker Ransomware. These findings provide valuable insights for developing effective countermeasures against ransomware attacks in the future. Making our dataset available open-source on Github . By that researchers can use it for future research and development in Ransomware Severity Prediction.

Autonomous Vehicle System (AVS) is rapidly advancing and is expected to completely transform the transportation industry, bringing about a new era of mobility. As digital data proliferation strains network resources, the demand for performing resource-intensive edge-assisted Deep Learning tasks in AVS with limited computational resources becomes increasingly challenging. The advent of 5G and future cellular networks (NextG) offers the promise of facilitating the seamless execution of these tasks. This research aims to diminish dataset size, in harmony with the integration capabilities of the Semantic and Flexible Open Radio Access Network (SEMO-RAN) framework, which is anticipated to reduce latency and refine resource allocation in vehicle image processing. The focus of this study is on optimizing image compression without compromising the accuracy of vehicle classification, leveraging esteemed Convolutional Neural Network models like YOLOv5 . By employing Generative Adversarial Network compression and Wavelet Image Compression methods, our study achieves an impressive 81.82% reduction in data size while maintaining a 96.97% accuracy in classification tasks. This underscores the potential for significant efficiency gains in AVS through improved data management, supported by Digital Twins (DT), Integrated Sensing and Communication (ISAC), and O-RAN.

Network Softwarization revolutionizes infrastructure by introducing programmable network functions, leading to scalable, agile, and efficient networks. This innovation facilitates adaptable, secure Next Generation Networks, transforming connectivity. This study conducts a comparative analysis of SDN-based network functions utilizing three different controllers: Open vSwitch (OVS), OpenFlow Reference controller, and RYU, to understand their impact on packet loss and jitter. The OpenFlow Reference controller, primarily used for educational and basic experimental purposes, serves as a baseline. The OVS controller, operating within the Open vSwitch framework in Mininet, offers a practical level of network management capability, more advanced than the OpenFlow Reference controller yet not as comprehensive as the RYU Controller, a production-grade SDN controller. Our research aims to fill the gap in comparative performance analysis of these controllers, particularly in hybrid round-robin-based load balancing and firewall implementations. Recognizing the operational reality of SDN deployments, we explore both reactive and proactive modes, acknowledging that real-world applications like firewalls and load balancers often operate proactively with pre-populated rules, affecting data plane performance. The findings provide insights into each controller’s effectiveness, with OVS and RYU showing enhanced capability in reducing packet loss and jitter compared to the OpenFlow Reference controller. This research not only sheds light on the functional behaviors and performance trade-offs of these controllers but also contributes to the broader SDN and NFV discourse, offering a foundation for future academic and practical advancements in the field.

The self-organizing method of distributed Wireless Sensor Networks (WSN) has an energy-efficient feature that can detect and assign channels to each node at different frequencies to avoid inter-node collision. The fading and interference of the channel may disconnect the wireless linkage between connected nodes. Most of the published work on this topic to date use digital filters to eliminate noise and interference for the wireless sensor network topology. However, estimation using a digital filter may not be the perfect candidate for communication in an underground mine, where node-to-node transmission is greatly affected by multipath fading. We propose a power control -based decision criteria for node transmission in case of deep fading to save energy. The and the node can also automatically re-establishment connection by adjusting its transmission power. The goal of the auto power control mechanism is to ensure the target Mean Squared Error performance in case of a random fading channel. The main objective of this work is to use a Random Walk -based Kalman Filtering scheme with Diffusion Least Mean Square (LMS) algorithm that can reduce the multi-path fading effect more accurately for successful signal transmission in in an underground environment. Our simulation results show that the proposed method outperforms the Mean Squared Error performance at low power consumption over the existing schemes.

In the digital age, the agricultural industry faces unique challenges, including the efficient management and maintenance of farm vehicles and tools. This paper introduces a blockchain-based agricultural vehicle and tool rental system leveraging Hyperledger Fabric. By utilizing blockchain technology, the system ensures immutability, transparency, and trust among all parties involved. The primary focus is on creating a decentralized and secure platform that not only facilitates the rental process but also introduces next-generation farm vehicle and tool maintenance capabilities. Throughout the development process, we faced and overcame significant challenges, including the need for a custom-tailored blockchain solution to meet the unique requirements of agricultural asset management. The proposed solution enhances traceability and accountability, thereby reducing fraudulent activities and improving overall operational efficiency. A working demo of the system is available on GitHub, providing an open-source resource for researchers and developers. This accessibility enables further exploration and adaptation of the system across various fields, from supply chain management to peer-to-peer marketplaces. The blockchain’s immutability aspect serves as a cornerstone for building trust in digital transactions, offering potential applications in diverse sectors such as finance, healthcare, and government services. The implementation details, innovative approaches to technical obstacles, and practical implications of our solutions are thoroughly discussed, providing valuable insights for researchers and practitioners in blockchain technology and agricultural systems.

Satellite image classification using Deep Learning (DL) algorithms is crucial for various applications such as Environmental Monitoring, Remote Sensing, and Urban Planning. The high-resolution nature of satellite images leads to large data volumes, resulting in computing challenges when classifying and transmitting the data over the internet. This paper addresses this challenge by reducing pixel size using Bicubic and Bilinear Interpolation. These techniques are known as image compression methods, which allow us to retain a significant portion of the original image information while reducing its size. Furthermore, adjusting the pixel size enables us to implement various levels of image compression supported by different DL classifiers, catering to diverse applications in NextG wireless networks or O-RAN. This pixel size reduction optimizes data transfer and speeds up satellite image transmission for Crop Monitoring and Land Cover Classification applications. After reducing the pixel, we employ different DL models such as CNN, ResNet, and stacked ensemble model, built by stacking three other EfficientNetB0 models to classify the low pixel images and compare the results using accuracy, precision, recall, and F1-Score. Staked EfficientNetB0 is the best performer among these, with an accuracy of 89.34% using a pixel size of 512x512. Our results show that we can reach up to 88.25% accuracy while maintaining a pixel size of 64x64, as our target is to reduce the image size while maintaining a tolerable accuracy so that network load and cost can be minimized.

The regular discovery of new attacks on 802.11 wireless devices emphasizes the importance of developing ML-based intrusion detection systems that generalize to such attacks. Class-imbalance issues in intrusion datasets pose a learning problem for ML solutions. Our methodology directly incorporates the class-imbalance issues to learn better feature importance measures. With features extracted using this technique from the AWID dataset, we use a gradient-boosted model to show that these features are necessary to generalize to new attack types in the AWID test dataset.

Traffic patterns in backbone IP networks often deviate from the norm, to lead to events commonly termed as anomalies. Different algorithms have been proposed in literature to identify anomalies, but very few classifiers have been proposed to classify and group the signalled anomalies. Moreover, the classification algorithms typically have a predefined number of classes and use supervised learning methods. Some classifiers apply the windowing method to make the large amount of data scalable into small groups. This paper proposes a novel method of classification of anomalous data packets with unsupervised learning using the technique of Correspondence Analysis. Correspondence Analysis does not need a predefined number of clusters to begin the classification and can handle a large amount of data. We have applied our developed algorithm on real data from the US Abilene backbone network, and compared our results with existing clustering algorithms. The results indicate that our proposed algorithm yields the best results in terms of classifying anomalies amongst the most recent classification algorithms available.

Distributed Denial of Service (DDoS) attacks continue to adversely affect internet-based services and applications. Various approaches have been proposed to detect different types of DDoS attacks. The computational and memory complexities of most algorithms, however prevent them from being employed in online manner. In this paper, we propose a novel victim-end online DDoS attack detection framework based on the celebrated Kernel-based Online Anomaly Detection (KOAD) algorithm and the Mahalanobis distance. We have employed the KOAD algorithm to adaptively model the normal behavior of network traffic, and then constructed the normal and abnormal datasets based on the results of KOAD. Subsequently, the Mahalanobis distance metric was calculated between datapoints of the abnormal and normal subsets. Finally, the chi-square test was used on the Mahalanobis distance values to segregate the DDoS attack datapoints from the normal ones. We have validated our algorithm on simulated DDoS scenarios, as well as real baseline data from a company operating in cyber security. Our results have revealed that our proposed hybrid approach boosts the performance of sole KOAD algorithm and Mahalanobis distance in detecting DDoS traffic in terms of both false positive and detection rates.

This paper presents a solution for Simultaneous localization and mapping (SLAM) with an autonomous differential drive robot in an indoor environment. We extract data from sonars and analyze it to construct a map of the explored area using occupancy grid mapping. In addition, relative positioning approaches provide location using an inertial measurement unit (IMU) and wheel encoders. Localization and mapping are essential tasks for an autonomous robot's navigation or exploration without a prior map and the system is based on discrete step-wise or event modeling which guides its navigation. The system's result is compared for accuracy against an actual map. The results show that the approach is largely accurate although the resolution of the map produced could be improved.

In this paper we propose a new algorithm based on kernel machines for automatic, instantaneous detection of emergencies occurring in a hospital Intensive Care Unit. The proposed algorithm takes as input the multitude of vital statistics that are continuously monitored for a critical patient in Intensive Care, learns the underlying pattern between the statistics that is naturally inherent for the particular patient, and instantaneously signals any deviation from this pattern. Through application to real data from a cardiac Intensive Care Unit at a hospital in a developing country, we show that it is possible to easily obtain high detection accuracy with low false alarm rates.

This paper considers the problem of predicting the number, length and distribution of IP traffic flows some time into the future, based upon packets collected in the present. A particle filter is used to predict the mean flow length and complete flow distributions for subsequent timesteps. A model for the histogram of flows corresponding to any given time interval is first presented, and the particle filter is then used to estimate the parameters of the model. The proposed method was tested on a large number of commonly-available data traces, and the results were analyzed in terms of the difference between the predicted flow distributions and actual flow histograms. An important application of this work is in resource reservation for protocols that require guaranteed qualities of service.

This paper considers the problem of predicting the number, length and distribution of IP traffic flows some time into the future, based upon packets collected in the present. Two versions of the Expectation-Maximization (EM) algorithm are used to predict the mean flow length and complete flow distributions for subsequent timesteps. A model is first used to represent the histogram of flows corresponding to any given time interval, and the EM algorithms are then used to estimate the parameters of the model. The proposed algorithms are tested on a large number of commonly-available data traces and both show high prediction accuracy.

In this paper we apply a recursive algorithm based on kernel mappings to propose an automated, real-time intruder detection mechanism for surveillance networks. Our proposed method is portable and adaptive, and does not require any expensive or sophisticated components. Through application to real images from BRAC University's closed-circuit television system and comparison with common methods based on Principle Component Analysis (PCA), we show that it is possible to obtain high detection accuracy with low complexity.

Large backbone networks are regularly affected by a range of anomalies. This paper presents an online anomaly detection algorithm based on Kernel Density Estimates. The proposed algorithm sequentially and adaptively learns the definition of normality in the given application, assumes no prior knowledge regarding the underlying distributions, and then detects anomalies subject to a user-set tolerance level for false alarms. Comparison with the existing methods of Geometric Entropy Minimization, Principal Component Analysis and OneClass Neighbor Machine demonstrates that the proposed method achieves superior performance with lower complexity.

High-speed backbones are regularly affected by various kinds of network anomalies, ranging from malicious attacks to harmless large data transfers. Different types of anomalies affect the network in different ways, and it is difficult to know a priori how a potential anomaly will exhibit itself in traffic statistics. In this paper we describe an online, sequential, anomaly detection algorithm, that is suitable for use with multivariate data. The proposed algorithm is based on the kernel version of the recursive least squares algorithm. It assumes no model for network traffic or anomalies, and constructs and adapts a dictionary of features that approximately spans the subspace of normal behaviour. The algorithm raises an alarm immediately upon encountering a deviation from the norm. Through comparison with existing block-based offline methods based upon Principal Component Analysis, we demonstrate that our online algorithm is equally effective but has much faster time-to-detection and lower computational complexity. We also explore minimum volume set approaches in identifying the region of normality.