Wildfire incidents have surged in frequency and severity in recent years highlighting the need for advanced technologies to predict wildfire behavior early and mitigate its impact. Recent strides in machine learning research, the increased availability of wildfire data, and computational resources have fueled the rise of data-driven approaches in wildfire management. This study aims to advance data-driven methods for predicting wildfire behavior and aid in timely decision-making and resource allocation efforts by adopting a Graph Neural Network (GNN)-based framework for predicting the burned area resulting from a wildfire ignition. GNNs have shown success in handling irregular-sized inputs and capturing the long-range dependencies inherent in geospatial data, such as wildfires, making them a viable alternative to CNNs which impose limitations on geospatial data due to their reliance on fixed-size inputs and local receptive fields. A framework is developed to represent spatial wildfire data and its influencing factors as graphs followed by the development of three distinct GNN models based on different message-passing mechanisms to process the graph-structured data. GNN models outperform CNN-based segmentation models in wildfire prediction, achieving higher AUPRC (0.4787), precision (0.4536), and AUROC (0.9377), and illustrating the efficacy of GNNs in modeling wildfire behavior by effectively capturing spatial dependencies.

In response to the growing incidence and severity of wildfires, this paper presents Vi-Net, a novel hybrid deep learning framework for next-day wildfire spread prediction. By integrating U-Net’s fine-grained spatial segmentation with the global contextual modeling of Vision Transformers (ViT), Vi-Net formulates wildfire spread prediction as a semantic segmentation task. The model is trained on a decade-long (2012–2020) multimodal wildfire dataset that integrates meteorological, topographical, and vegetation features. To address the severe class imbalance inherent in wildfire data, Vi-Net employs a Focal Tversky loss function. Experimental results show that Vi-Net achieves an F1-score of ∼97% and an Intersection over Union (IoU) of ∼94% on test data, significantly outperforming standalone U-Net and ViT models. These findings underscore Vi-Net’s potential to improve wildfire mitigation planning, resource allocation, and emergency response.

Forest fires or wildfires are becoming more prevalent across Canada. They are both beneficial and harmful. They promote forest health and aid ecological processes. However, they play a devastating role in impacting the economy of a nation and also impact the health of humans. Hence, it is important to consider all data sources relevant to forest fires or wildfires. The Canadian Wildland Fire Information System (CWFIS) calculates the danger of forest fires. The Canadian Forest Fire Weather Index (FWI) System is a critical part of CWFIS, which does not consider land vegetation in its calculations. Considering it is the vegetation that burns in a forest fire, it is important to have an insight into what types of vegetation are more prone to fires. Earth observation data for vegetation over land is now available across North America. This research primarily provides an analytical data visualization of the vegetation land cover impacted by and impacting forest fires. We look into open-source vegetation land cover data and provide insights into forest fires or wildfires. A look into the change of vegetation is also provided.

The advancement of big data has paved the way for the development of intelligent and smart applications; however, privacy concerns often hinder fully realizing their benefits. Federated Learning (FL) has emerged as a promising framework for enhancing privacy while training models collaboratively across decentralized data sources. However, it remains susceptible to poisoning attacks, severely undermining its effectiveness. Existing robust aggregation techniques often struggle with the sensitivity of data distributions, and cluster-based strategies often fail to cluster poisoned model updates correctly. The direction obtained from the signs of the gradient mostly solves these problems but remains vulnerable to dying ReLU problems and usually becomes sensitive to outliers. In this paper, we introduce SignDefence, a sign direction and LeakyReLU-based aggregation technique, which considers the direction of the gradients and overcomes the performance issues related to the dying ReLU problem. Moreover, the proposed SignDefence computes Jaccard Similarity over binary encoded model weights and remains robust across sparse data. The experimental results suggest that the proposed technique shows consistently better accuracy and F1 score than the state-of-the-art techniques, without attack and under different attack scenarios.

This paper proposes the Server–Client Machine Learning Intrusion Detection System (SC-MLIDS), a novel fusion framework designed to enhance security in Wireless Sensor Networks (WSNs), which are inherently vulnerable to various security threats due to their distributed nature and resource constraints . Traditional Intrusion Detection Systems (IDSs) often face challenges with high computational demands and privacy issues. SC-MLIDS addresses these problems by integrating Federated Learning (FL) with a multi-sensor fusion approach to implementing two layers of defence that operate independently of specific attack types. Moreover, this framework leverages a server–client architecture to efficiently manage and process data from sensor nodes , sink nodes, and gateways within the network. The core innovation of SC-MLIDS lies in its dual model aggregation algorithms at the gateway: one assesses model performance and weight, while the other uses majority voting to integrate predictions from both client and server models. As a result, this approach reduces redundant data transmissions and enhances detection accuracy, making it more effective than conventional methods in WSNs. Our proposed framework outperforms current state-of-the-art techniques, achieving F1-scores of 99.78% and 98.80% for the two aggregation algorithms, namely, Weighted Score and Majority Voting. This validation demonstrates the effectiveness of SC-MLIDS in providing accurate intrusion detection and robust data management.

Prevention of forest fires increasingly impacted by climate change is essential to maintain ecological balance, preserve natural resources, prevent economic loss, and protect human and animal life. Data for forest fires is available from multiple sources and is huge. Federated learning can be implemented to distribute the computing across multiple edge devices by saving transmission costs, protecting data privacy, and maintaining security with no single point of failure as local models exist across multiple resources in different geographic regions. The proposed framework extends the Spatio-Temporal Agnostic Subsampling (STAS) technique by distributing the data into multiple computation nodes to leverage federated learning. It was found that the models trained using federated learning on weather data gained on average 0.3 in F1 for classifying the occurrence of fire. This study also demonstrates how to optimally choose the sources of data for either predicting the occurrence of fire or the severity of fire.

Federated Learning (FL) has emerged as a novel distributed Machine Learning (ML) approach, to tackle the challenges associated with data privacy and overload in MLbased intrusion detection systems (IDSs). Drawing inspiration from the FL architecture, we have introduced a hybrid ML IDS tailored for Wireless Sensor Networks (WSNs). This system is crafted to leverage ML for achieving a two-layer intrusion detection mechanism in WSNs free from constraints posed by specific attack types. The architecture follows a server-client model compatible with the configuration of sensor nodes, sink nodes, and gateways in WSNs. In this setup, client models located at sink nodes undergo training using sensing data while the server model at the gateway is trained using network traffic data. This two-layer training approach amplifies the efficiency of intrusion detection and ensures comprehensive network coverage. The results derived from our simulation experiments corroborate the effectiveness of the proposed hybrid ML IDS. It generates precise aggregation predictions and leads to a substantial reduction in redundant data transmissions. Furthermore, the system exhibits efficacy in detecting intrusions through a dual validation process.

Supervisory Control and Data Acquisition (SCADA) systems are vulnerable to traditional cyber-attacks, such as man-in-the-middle, denial of service, eavesdropping, and masquerade attacks, as well as future attacks based on Grover's and Shor's algorithm implemented in quantum hardware. This paper proposes a quantum-robust scheme based on entanglement and supersingular isogeny-based cryptography. The scheme employs a modified Supersingular Isogeny Key Encapsulation (SIKE) to generate shared secret keys, also authenticating BBM92, a quantum key distribution protocol to generate a symmetric key. The paper uses ASCON-128 and SHA-3 to encrypt and authenticate messages, and provides a comparative analysis of two entanglement-based quantum key distribution protocols. The proposed scheme is compared to the current SCADA standard, AGA-12, and is shown to provide confidentiality, integrity, intrusion resistance, message authentication, and scalability. The randomness of key pairs generated by our algorithm and RSA key pairs is 87.5% and 84.37%, respectively, addressing confidentiality and integrity. Using the BBM92 protocol, our proposed algorithm detects the presence of an adversary by generating an average error rate of 26.07% and information leakage of 76.01%. AGA-12 relies on SHA-1 hash function that Google has cracked recently. However, our algorithm includes SHA-3, a collision and quantum-resistant hash that provides message authentication.

Forest fires pose imminent threats to ecosystems and human lives, necessitating precise prediction for effective mitigation. The challenges include managing extensive big data and addressing data imbalance. This study introduces a data integration framework that integrates data from remote sensing satellites, ground-based weather stations, and other sources to create a comprehensive weather database spanning 18 years in Alberta, Canada. Machine learning methods, including Random Forest, eXtreme Gradient Boosting, and Multi-Layer Perceptron are employed to evaluate forest fire prediction performance, overcoming the challenge of data imbalance through changes in spatial resolution, spatio-subsamping, and downsampling techniques. XGBoost exhibits results with an ROC-AUC score of 87.2% and a sensitivity of 75%.Using meteorological data and fire history improves prediction, demonstrating big data and machine learning’s role in addressing forest fire challenges.

Fake news has become a major concern due to its spread on social media. To combat this, various machine learning (ML) techniques have been proposed. However, there is a lack of research on the performance of transformer models using datasets from a wide range of domains. This paper investigates the performance of ML algorithms on three fake news datasets: LIAR, FNC-1 and Balanced Dataset for Fake News Analysis. Pretrained transformer language models such as BERT, RoBERTa, ALBERT and DistilBERT were chosen for this paper. The performance of the models was consistent across all datasets. RoBERTa obtained an accuracy of 69% when trained on the LIAR dataset, an 11% improvement over the existing traditional and deep learning ML model implementations, and an accuracy of 97% when trained on the FNC-1 dataset, proving to be the best-performing model across all the fake news detection datasets utilized in the experiments. DistilBERT trains at a significantly faster rate than the other three variants. The experimental results from the paper can help the research community to continue investigating and gain insights into fake news detection.

This research provides a spatio-temporal agnostic framework based on subsampling to generate generic deep learning models using publicly available weather data and to predict the probability of forest fire and severity. The aim is to show that this framework can be used to subsample and generate a balanced dataset for generic deep learning models to improve predictions for forest fires. The framework works for binary classification and regression deep learning models. It also works with limited variations between fire and non-fire data. Using this framework, 45 of the binary classification models built produced an F1Score greater than 0.95 while 35 of 54 regression models produced an R2Score greater than 0.91.

The proliferation of technologies and unstructured data on the internet poses a persistent challenge in extracting valuable information from diverse formats. To address this, research leverages Machine Learning (ML) and Natural Language Processing (NLP) techniques. This study contributes to information extraction from unstructured text using a state-of-the-art pipeline, incorporating modules for coreference resolution (Neuralcoref), named entity linking (Wikifier API), and Relationship Extraction (RE) (OpenNRE and REBEL models). The resulting Knowledge Graph (KG) in Neo4j captures entity relationships. Experiments on a BBC news dataset analyzed the pipeline’s performance, focusing on RE. Accuracies of 61.4% (OpenNRE) and 87% (REBEL) were achieved. The research demonstrates the efficacy of the proposed pipeline in extracting structured knowledge from unstructured data, facilitating the preservation and utilization of valuable information.

Federated Learning (FL) is an emerging paradigm that enables multiple clients to train a global model collaboratively without sharing their privacy-sensitive data. However, one of the significant challenges in FL is the aggregation of the model updates from different client devices, as malicious participants acting as Byzantine attackers can craft the model update and poison the global model. The state-of-the-art defence mechanisms mostly rely on aggregation-based security defences to improve the degraded accuracy. However, preventing attacker's participation in the training can have an impact on improving the global model's accuracy. Therefore, in this paper, FedChallenger, a dual-layer defence mechanism, is proposed, which attempts to detect and prevent malicious participation in the FL training process in its first layer. The other layer incorporates a trimmed-mean aggregation strategy, where pairwise cosine similarity identifies malicious updates and removes entire client updates from federated averaging. Extensive experiments using the BloodMNIST dataset validate that the FedChallenger gains nearly 85%, 80%, 15%, and 4% accuracy with more than 1.2 times faster convergence rate over the state-of-the-art Byzantine resilient aggregation strategies called FedAvg, Fang, Krum, and Trimmed-Mean approach, respectively, on 40% compromised devices. Above all, it shows consistently better results than them in both attack and non-attack scenarios.

Cyber-security experts often require the help of an automated process that filters and classifies network attacks. To apply specific preventive measures for securing networks, the classification of the attack type is the key. Many Machine Learning (ML) models have been proposed as a base for Network Intrusion Detection (NID) systems. However, their performance varies based on multiple factors. For instance, an ML model fitted on a highly imbalanced dataset can be biased toward over-represented attack types. On the other hand, paying attention only to the ML model’s performance in the minority classes can negatively affect its performance in the majority classes. This paper proposes an NID system that addresses the issue of imbalanced datasets and uses Convolutional Neural Networks (CNN) to classify the different attack types. We compare the performance of our proposed system to other systems that use: Random Over-Sampling (ROS), Synthetic Minority Oversampling TEchnique (SMOTE), Adaptive Synthetic Sampling (ADASYN), and Generative Adversarial Networks (GAN). Using the NSL-KDD and the BoT-IoT datasets for benchmarking, we show that our proposed system performs well in the minority classes: recall scores of 70.50% and 72.08% on the User to Root (U2R) and Remote to Local (R2L) attack classes of the NSL-KDD dataset, respectively, while maintaining an overall False Alarm Rate (FAR) of 6.50% and a recall of 90.46% on the binary classification task. Our proposed system scores a weighted average F1-Score of 99.45% on the multi-class classification task using the BoT-IoT dataset.

Supervisory Control and Data Acquisition (SCADA) networks play a vital role in industrial control systems. Industrial organizations perform operations remotely through SCADA systems to accelerate their processes. However, this enhancement in network capabilities comes at the cost of exposing the systems to cyber-attacks. Consequently, effective solutions are required to secure industrial infrastructure as cyber-attacks on SCADA systems can have severe financial and/or safety implications. Moreover, SCADA field devices are equipped with microcontrollers for processing information and have limited computational power and resources. This makes the deployment of sophisticated security features challenging. As a result, effective lightweight cryptography solutions are needed to strengthen the security of industrial plants against cyber threats. In this paper, we have proposed a multi-layered framework by combining both symmetric and asymmetric key cryptographic techniques to ensure high availability, integrity, confidentiality, authentication and scalability. Further, an efficient session key management mechanism is proposed by merging random number generation with a hashed message authentication code. Moreover, for each session, we have introduced three symmetric key cryptography techniques based on the concept of Vernam cipher and a pre-shared session key, namely, random prime number generator, prime counter, and hash chaining. The proposed scheme satisfies the SCADA requirements of real-time request response mechanism by supporting broadcast, multicast, and point to point communication.

Smart grids rely on SCADA (Supervisory Control and Data Acquisition) systems to monitor and control complex electrical networks in order to provide reliable energy to homes and industries. However, the increased inter-connectivity and remote accessibility of SCADA systems expose them to cyber attacks. As a consequence, developing effective security mechanisms is a priority in order to protect the network from internal and external attacks. We propose an integrated framework for an Intrusion Detection System (IDS) for smart grids which combines feature engineering-based preprocessing with machine learning classifiers. Whilst most of the machine learning techniques fine-tune the hyper-parameters to improve the detection rate, our approach focuses on selecting the most promising features of the dataset using Gradient Boosting Feature Selection (GBFS) before applying the classification algorithm, a combination which improves not only the detection rate but also the execution speed. GBFS uses the Weighted Feature Importance (WFI) extraction technique to reduce the complexity of classifiers. We implement and evaluate various decision-tree based machine learning techniques after obtaining the most promising features of the power grid dataset through a GBFS module, and show that this approach optimizes the False Positive Rate (FPR) and the execution time.

In this paper the technology of collecting the logs and the logs message format is considered in addition to investigating multiple log data sources as an input to the cloud management system. A comparison between message exchange technologies (JSON, XML) is evaluated with the latest message format technology (Google Protocol Buffer) when used in combination with the message transmission protocols (XML-RPC, REST, Network Socket). In addition, the sampling rate that gives the accurate reading of resource usage is investigated, which is used to select among different log data sources to achieve the accurate log update time. Logs sampling rate of 1.0 second is found to be the best with "xentop" data source. The result of the experiment shows using Protocol Buffer with Socket protocol gives the best results in reducing message size. Network socket with JSON gives the best processing delay and traveling time.

Cloud computing model is the most complex computing model that requires implementing effective techniques to manage infrastructure resources of datacenters. Unproductive tasks scheduling can lead to an increase in the operational cost of cloud provider side, which in turn increases the cloud services cost at cloud consumer side. One of the effective techniques to address these issues in cloud datacenters is the elasticity by allowing dynamic resource provisioning based on the current demand and varying workload running upon virtual machines (VMs) over time. This leads to an increase in the resource utilization, and reduced power consumption by turning off the idle physical machines. However, the dynamic resource provisioning due to the growing service demand and higher quality of service requirements of the users can cause a violation of service level agreement. In this paper, we propose a model based on linear regression to manage and reformulate cloud users requests and dynamically generating rules based on historical data of their requests in order to update association functions to address and adapt the changes of different types of workloads running on the cloud provider datacenter. The experiments and simulation results based on dynamic workloads show the proposed algorithm significantly increases the resource utilization on cloud datacenter.

Network traffic anomaly may indicate a possible intrusion in the network and therefore anomaly detection is important to detect and prevent the security attacks. The early research work in this area and commercially available Intrusion Detection Systems (IDS) are mostly signature-based. The problem of signature based method is that the database signature needs to be updated as new attack signatures become available and therefore it is not suitable for the real-time network anomaly detection. The recent trend in anomaly detection is based on machine learning classification techniques. We apply seven different machine learning techniques with information entropy calculation to Kyoto 2006+ data set and evaluate the performance of these techniques. Our findings show that, for this particular data set, most machine learning techniques provide higher than 90% precision, recall and accuracy. However, using area under the Receiver Operating Curve (ROC) metric, we find that Radial Basis Function (RBF) performs the best among the seven algorithms studied in this work.

Live VM migration is a technique that consists of a selection process and a migration process to migrate a VM from one host to another in the same data center without changing the IP address, or in a different data center with necessity to the VM to get a new IP address. The changing of IP address results into a mobility problem, which may render the service unreachable. In this paper, we propose a system model that selects data center randomly for VM placement while reducing this IP address reconfiguration. A new metric is proposed to indicate number of users that need IP reconfiguration. We extended CloudSim to simulate our work to identify the number of IP reconfigurations required for VM migration across the data centers on random workload.

Different applications in smartphones result in different power consumption patterns. The fact that every application has been coded to perform different tasks leads to the claim that every action onboard (whether software or hardware) will consequently have a trace in the power consumption of the smartphone. Even though the power consumed by the application might not be the same every time it is used, there still remains a similarity in the power consumption pattern. An anomalous behavior on the smartphone would result in a reduction in the similarity of the power consumption pattern. This change in similarity can be used to detect the presence of anomalous behavior of smartphones. We have proposed two approaches to detecting anomalous behavior on smartphones based on the power consumption pattern. The first approach is based on signal processing and the second approach explores the area of statistical learning in detecting malware. The two approaches have been analysed, evaluated, and compared. It has been observed that the signal processing method of detection performed better for anomalous behavior of lower intensity and the statistical learning method performed better for higher intensity anomalous behavior. It was also observed that both the methods are complementary.

A private cloud is maintained by an enterprise forits internal use. In such a scenario instead of buying the resources the enterprise can acquire the resources from a public cloud such as the ones provided by Amazon and Microsoft. On conventional systems rigorous analysis of the system and its workload is performed for determining the appropriate number of resources to be deployed on the private cloud. This paper presents a middleware framework that avoids this step of a priori capacity analysis and allows such private cloud owners to provision resources automatically such that a specified grade of service is maintained. The proposed framework performs dynamic resource provisioning that also leads to a reduction of operational cost. Additional resources are acquired during high traffic periods and released during low traffic periods such that the desired grade of service is always maintained. The paper describes the architecture of the framework and the experience gained from a prototype implementation including a preliminary analysis of its performance.

Software architectural design has an enormous effect on downstream software artifacts. Decomposition of function for the final system is one of the critical steps in software architectural design. The process of decomposition is typically conducted by designers based on their intuition and past experiences, which may not be robust sometimes. This paper presents a study of applying the clustering technique to support system decomposition based on requirements and their attributes. The approach can support the architectural design process by grouping closely related requirements to form a subsystem or module. In this paper, we demonstrate our experiments in applying the approach to an industrial communication protocol software system and comparing several clustering algorithms. The result obtained from WPGMA (weighted pair-group method using arithmetic averages) shows closer resemblance than other clustering methods to the one developed by the designer.

This report presents a reflection on software architecture practices based on our past ten year’s industrial experiences, particularly in the area of telecommunications. The report summarizes the methods, tools, and techniques that we have used on various projects. We also discuss, based on our experiences, what methods are useful, what remains to be validated, and what the gaps are between the state of practices and our wishes.