Workflows have emerged as the dominant methodology in application development due to their flexibility in building a wide range of applications, including multi-tier web applications, scientific computing applications, and big data applications. The workflow plays a crucial role in orchestrating the distributed services that constitute an application. Containers emerge as the most favourable lightweight virtualisation technique for isolating, encapsulating, and deploying services, thanks to their superior portability when compared to Virtual Machines (VMs). While container-based systems offer various advantages, the integration of these systems into workflow scheduling remains one of the most formidable challenges in distributed environments. Indeed, significant issues arise due to the growing demands of clients. In this survey paper, we present a Systematic Literature Review (SLR) on the current state of the art in this domain. We propose a taxonomy to compare and evaluate the existing studies on workflow scheduling approaches for container-based systems. This taxonomy encompasses various criteria, including scheduling techniques, performance metrics, and container management systems. We highlight certain recommendations for open issues which require more investigations. Our aim is to provide valuable insights for researchers and developers interested in understanding the contributions and challenges of current workflow scheduling approaches for container-based systems. This knowledge can serve as a foundation for enhancing their solutions.
The proliferation of IoT devices due to cyber threats, which have become increasingly sophisticated, requires a strong security framework. This paper proposes a new framework for Intrusion Detection System-IoT-IDs using a Random Forest classifier to first classify the attack into binary features and prepare a new data set that would enable multiclass classification. It achieved an overall accuracy of 0.98 on the comprehensive UNSW-NB15 data set, with very good performance in detecting 'Generic' attacks, having almost perfect precision, recall and F1-score. It also presents cases of 'Analysis' and 'Backdoor' types of attacks, where further improvements should be done. All these models have been analysed to find the pros and cons in IoT settings using Random Forests, XGBoost and MLP. Further studies based on the research could be done on multiplying models with improved features, intrusion detection in real time and more strong AI techniques. This paper focuses on addressing challenges with imbalanced classes and scalability concerns using data privacy preservation methods for improving the performance of IDS. It represents one step further in continuous improvement with respect to security and reliability of IoT networking, while at the same time opening wide avenues for future research and advances in IDS technologies.
In recent years, smart grids have facilitated the integration of renewable energy sources (like solar and wind) and energy storage systems using Internet of Things (IoT) devices and smart applications. This integration helps balance supply and demand and improves grid resilience for Intelligent Electronic Devices (IEDs). Deploying the IEC-61850 standard for communication between Intelligent Electronic Devices (IEDs) indeed introduces new security challenges due to its specific architecture and communication protocols tailored for smart grids. With increased digital connectivity, smart grids implement robust cybersecurity measures to protect against cyber threats and ensure data privacy. Cybersecurity in smart grids and IoT is crucial due to the increasing digitalisation and connectivity of critical infrastructure. On the other hand, Machine Learning (ML) can play a crucial role in cybersecurity analysis of IEDs such as Intrusion Detection Systems (IDS) in smart grids by leveraging data analytics and pattern recognition techniques. In this paper, a Genetic-based Ensemble Forest Algorithm (GEFA) is presented to predict the attack surface for cyber threats of IEDs in IoT-based smart grids. Also, a feature selection method based on Ant Colony Optimisation (ACO) algorithm is applied for a real public power system data set to enhance the performance of prediction procedure. The experimental results show that the suggested hybrid ACO-GETA approach outperforms other prediction approaches to achieve highest accuracy and F1-Score with 100%.
This paper presents an adaptation of online communities to manage crises effectively. With the COVID-19 pandemic forcing many activities to move to the online environment, online communities have become increasingly vulnerable to crises. The uncertainty surrounding crises poses significant project management risks in various areas. Therefore, developing effective strategies for crisis adaptation in online communities is crucial. The authors propose a decision-making framework for managing online university communities to mitigate risks. This approach was implemented in six university online communities. The algorithm was developed to adapt online communities to a crisis and prevent them from falling into risks or crises. The algorithm considers factors generated by the COVID-19 crisis. The algorithm results are recorded and submitted to the project manager and stored in the project documentation database for permanent analysis of the university's online community. This approach allows online communities to manage crises effectively and reduce the associated risks. This study emphasises the importance of adapting online communities to the risks associated with managing crises. It provides a practical framework for decision-making and a well-designed algorithm that can prevent online communities from falling into risks or crises. This research has significant implications for managing crises in the online community and is used by universities and other organisations.
For safe driving and decreasing the risk of traffic accidents, it is important to control the driver's mental status. In this paper, an intelligent system is implemented based on Fuzzy Logic (FL) for deciding Driver Mental Status (DMS). In order to investigate the effects of the considered parameters we implement two models: DMS Model 1 (DMSM1) and DMS Model 2 (DMSM2). The input parameters of DMSM1 include Driver Anxiety Level (DAL), Traffic Situation (TS), Driving Operating Time (DOT), while for DMSM2 we add a new parameter called Drive Distress Situation (DDS). For both models, the output parameter is DMS. We compared the simulation results of DMSM1 and DMSM2. The evaluation results show that DMSM2 is more complex because the rule base is bigger than DMSM1, but it has a better decision of DMS value.
As attacks on the network environment are rapidly becoming more sophisticated and intelligent in recent years, the limitations of the existing signature-based intrusion detection system are becoming more evident. For new attacks such as Advanced Persistent Threats (APT), the signature pattern has a problem of poor generalisation performance. Research on intrusion detection systems based on machine learning is being actively conducted to solve this problem. However, in the actual network environment, the attack sample is collected less than the normal sample, so it suffers a class imbalance problem. When a supervised learning-based anomaly detection model is trained with these data, the results are biased toward normal samples. In this paper, AutoEncoder (AE) is used to perform single-class anomaly detection to solve this imbalance problem. The experimental evaluation was conducted using the CIC-IDS2017 dataset, and the performance of the proposed method was compared with supervised models to evaluate the performance.
In the era of the fast development of data technologies, there is an urgency to continuously elevate the level of standardisation and legislation in building a social credit system since credit default behaviour is still widespread across society. It is critical to evaluate and classify citizens' credit risk with the help of domain experts. It can provide an effective basis for public security organs. Because of its well-known stability and interpretability, this study employs logistic regression for the analysis of the data and propose pertinent crime prevention and control measures. A logistic regression model is built and applied to predict and assess individuals' credit risk, and the model's prediction accuracy is 76.7%. Next, K-means clustering is used to divide the scores into different clusters to facilitate the grading of credit default risks of individuals. Based on our method, the relevant authorities can impose hierarchical prevention and management for ex-offenders and key personnel, which provides new, robust support for intelligent governance.
Wordle is a popular daily puzzle in the New York Times. The 'Predicting Wordle Results' problem in the 2023 Mathematical Contest in Modelling (MCM) focused on developing a model to estimate the reported headcount in the difficult mode. The model considered three attributes: word frequency, letter repetition and letter frequency. The analysis showed that these attributes influenced the reported headcount. Correlation analysis revealed a strong relationship between the number of participants and word frequency and letter repetition. A neural network time series model was developed, using word frequency and letter frequency as inputs to predict the reported results. The model achieved a high accuracy with an R & sup2;-value of 0.95. The study found that the number of participants had a linear relationship with the number of participants in the difficult mode. Most word frequencies in the questions were below 0.0002.
We present the design and implementation of a web page that allows access to private information, such as the availability of professors in their office, only when it is accessed from a specific location (e.g., entrance hall of a building on a university campus). The web page initially contains no private information; instead, it contains some buttons to request it. When one of the buttons is tapped by a user, a JavaScript function acquires a short-time password broadcast by a Bluetooth Low Energy beacon installed at the location and sends it to the web server. If the password is valid, the web server returns the private information, which is used to replace the button labels. Access to the private information is thus limited to those present at the specific location. The ability to check the availability of professors in advance will save students and visitors time and effort.
Typically, universities aim to achieve a high position in ranking systems for their reputation. However, self-evaluating rankings could be costly because the indicators are not only from bibliometrics, but also the results of over a thousand surveys. In this paper, we propose a novel approach to estimate university rankings based on traditional data, i.e., bibliometrics, and non-traditional data, i.e., Altmetric Attention Score, and Sustainable Development Goals indicators. Our approach estimates subject-areas rankings in Arts & Humanities, Engineering & Technology, Life Sciences & Medicine, Natural Sciences, and Social Sciences & Management. Then, by using Spearman rank-order correlation and overlapping rate, our results are evaluated by comparing with the QS subject ranking. From the result, our approach, particularly the top-10 ranking, performed estimating effectively and then could assist stakeholders in estimating the university's position when the survey is not available.
The increase in demand for reducing the latency in service requests of Internet of Things (IoT) applications has led researchers to drift from Cloud computing to Fog computing paradigms. Fog computing brings computing and data storage closer to devices and sensors, reducing latency and improving response time and reliability. However, implementing fog computing successfully requires fog nodes and applications to be deployed effectively to provide high-performance services. In addition, fog computing faces challenges in efficient resource scheduling due to the scarce capacity of fog nodes and the dynamic and heterogeneous nature of devices, leading to complexities in workload allocation and optimal resource utilisation. This work presents a simplified model for dynamically placing the application modules in a heterogeneous fog computing environment. A new framework for learning scheme is implemented using a Deep Deterministic Policy Gradient (DDPG)-based reinforcement learning technique for predicting the operations and determining the cumulative rewards. A test environment demonstrates that the proposed framework has lower mobility dependency, higher reward and reduced variance compared to existing schemes.
IoT applications have actively employed advanced technologies, utilising neural networks to recognize, analyse, and interact with their surroundings. Significantly, Amazon Echo exemplifies an IoT application by bridging physical and human realms with the digital domain, employing deep learning for voice command comprehension. Similarly, Microsoft's Windows facial recognition security system integrates DL to unlock doors upon facial recognition. This research explores the integration of IoT with DL techniques to enhance the monitoring and analysis of drinking water quality assessment, evaluating various methods to determine drinkability. Various machine learning algorithms, including Random Forest, LightGBM and Bagging Classifier, are employed to predict water quality based on multiple parameters such as pH, conductivity and turbidity. The study uses a comprehensive data set featuring 3277 values for nine different water quality indicators. Comparative analysis revealed similar outcomes: Random Forest demonstrated the highest accuracy, achieving a predictive accuracy of 0.824695 followed by Light GBM and Bagging Classifier. This research contributes to the ongoing efforts to employ advanced computational techniques in environmental monitoring, providing a reliable methodological framework for future studies to enhance water quality assessment.
VM consolidation is an effective approach for reducing energy consumption in cloud data centres. The selection of VMs from under-loaded or overloaded machines and migrating them on effective hosts constitutes the process of VM consolidation. The algorithms in literature to select hosts for VM deployment are generally based on a single criterion. However, VM placement is a multi-criteria decision-making problem. In this paper, an attempt is made to design a host selection technique based on improved Preference Ranking Organisation Method for Enrichment Evaluations (PROMETHEE) for energy efficient VM deployment. The proposed selection policy uses multiple parameters to find the selection index of hosts. The selected hosts help to reduce both energy consumption and service level agreement violations. A case study-based approach is followed to validate the proposed VM deployment framework using real data, real hosts and VMs configuration. Results indicate the employability of the framework in real cloud environments.
To address security threats to cloud server serial port communication data, the research introduces a monitoring model that incorporates a passive clustering algorithm. The study uses a density-based approach to identify natural clusters in the data, allowing each cluster to have a different shape and size. The algorithm first evaluates the local densities of the data points and then assigns the data points to the nearest high-density areas based on these density values to form clusters. The results revealed an improvement of up to 0.06% in recall, up to 0.12% in accuracy and up to 0.09% in F1-score. The passive clustering algorithm improved on an average of 32.56% in F1-score, 24.78% in accuracy and 3.38% in recall compared to other methods. More advanced optimisations further improved the detection accuracy to 98.5678%, the false alarm rate to 1.4322% and the detection latency further reduced to 16,888.43 ms, highlighting the potential of the passive clustering algorithm in monitoring the security of serial port data for cloud server communication. As a result, the model constructed by the research can optimise the limitations existing in the traditional sub-cluster model, and then realise the service of user information data security, which has excellent practical value and broad application prospects.
