This study demonstrates how fake news can be detected and analyzed by diffusion trends, mainly by treating it either as a classification problem using ML (Machine Learning) or by examining how it spreads on $\operatorname{Twitter}(\mathrm{X})$ temporal network. Classification involves supervised learning, where labeled training data is used to build predictive models. In this comparative analysis, we evaluate the effectiveness of four wellknown ML algorithms-K-Nearest Neighbors (K-NN), Naive Bayes, Random Forest and BigBird-in detecting fake news. The goal is to compare their accuracy in classifying textual tweet text into two categories: real and fake news. The WELFake dataset, which contains a variety of Twitter-based text, is used for experimentation. After identifying BigBird as the most effective model for our dataset, we apply it to a new tweet text related to COVID-19 news to determine its classification. The models were then tested on a tweet related to COVID-19, and each one identified the content as likely being fake news. In the final phase of the study, we applied two epidemiological models: SIR (Susceptible, Infected, Recovered) and SEIZ (Susceptible, Exposed, Infected, Skeptics) to compare the spread dynamics of contrasting viewpoints. It was observed that the SEIZ model fits the data more accurately and with less error than the SIR model. Furthermore, the $\mathbf{2 4}$-hour prediction analysis of infected nodes using the SEIZ model demonstrates that the diffusion of fake news follows a typical viral spread pattern, characterized by a rapid initial increase in infected nodes, followed by a plateau as the network reaches saturation.
Energy efficiency is becoming increasingly important for sustainability. In a cloud-native environment, orchestration platforms such as Kubernetes can significantly impact the overall energy performance of applications as they manage their lifecycle, including scheduling. Kubernetes, specifically as a de facto cloud platform, has a default scheduler that does not account for energy consumption in its decision-making process, highlighting the need to explore mechanisms that can improve the energy performance of deployed applications. In this paper, we present an energy-aware scheduling mechanism within Kubernetes to optimize the energy usage of cloud-native applications while meeting other scheduling requirements and constraints. This approach extends the scheduling logic in Kubernetes through plugins that integrate energy metrics and employ Machine Learning-based power estimation for energy-optimized decision-making. The evaluation of the proposed approach shows up to 12 % energy savings with minimal performance degradation. When complemented by other consolidation techniques, it could potentially lead to substantial savings ranging from 21 % to 65 %, depending on the system load in the scenario analyzed. Importantly, this method does not introduce significant overhead.
Research in the Network-based Intrusion Detection Systems (NIDS) field is a major focus for academics and the emergence of reliable and applicable detection models quickly became imperative for the integrity of critical infrastructures. Thus, Machine Learning (ML) models-specifically Deep Learning (DL) models- have been developed to detect network attacks and multiple training datasets are available, which allows to reproduce experiments. Nevertheless, the problem of intrusion detection is challenging since generating a representative set of training data is difficult due to the large variety of applications, and the fact that network infrastructures evolve over time. Also, a key concern often ignored is the different performances of models between experiments on known and often curated datasets, and in reality when deployed. In this paper, we propose a methodology to highlight the shortcomings of Anomaly-based NIDS, showcasing significant performance drops using realistic scenarios of evolution in the datasets. We implement a Denoising AutoEncoder (DAE) as a NIDS model. Then, we present a method to test and interpret the NIDS results based on analyzing the DAE's Mean Squared Error (MSE) to provide insights into the model's flow classifications. We advocate for considering the natural evolution of infrastructures when designing NIDS and highlight further research to address this challenge.
Optical turbulence is a phenomenon caused by atmospheric temperature variations and wind shear, leading to changes in air refractive index. This affects light propagation through the atmosphere, resulting in wavefront distortion, scintillation, and image degradation, posing a significant challenge to the reliability of free-space optical (FSO) communication systems. This paper presents a novel ensemble machine learning framework for predicting the refractive index structure parameter ($C_{n}^{2}$) in operational FSO networks. By integrating multiple machine learning algorithms (Random Forest (RF), Neural Network (NN), and Support Vector Machine (SVM), the proposed approach is designed to perform effectively under dynamic environmental conditions. The ensemble model achieves superior accuracy (97.82%) as shown in Fig. 3 and operational robustness compared to conventional methods. Through comprehensive uncertainty quantification, we identify key atmospheric variables influencing prediction confidence, with wind speed as the strongest predictor ($r=0.770$), followed by temperature gradient ($r=0.306$), as seen in Fig. 5. Our method reduces FSO link outages by approximately 20 % relative to traditional techniques through condition-specific uncertainty modeling that enables preemptive adjustments before link degradation. This framework provides a practical strategy for real-time optimization of optical systems across varying atmospheric conditions, particularly benefiting deployments in high-wind scenarios and complex terrain environments.
Large-scale learning management systems (LMS) generate rich interaction logs that offer valuable insights into study behavior over time. This paper analyses more than 11.3 million activity records collected from $\mathbf{5, 3 0 0}$ students enrolled in 1,166 Moodle courses during four consecutive semesters between winter $2022 / 23$ and summer $2023 / 24$. We replicate and extend our 2023 baseline study by (i) contrasting activity volumes, (ii) comparing hourly access patterns using two-sample KolmogorovSmirnov tests, and (iii) testing grade and time-to-deadline distributions with Mann-Whitney and Wilcoxon statistics. While the temporal shape of weekly access curves remained remarkably stable across cohorts ($p > 0.16$ for all KS tests), the absolute number of interactions grew by 60 %. At the same time, students in the most recent summer term submitted assignments significantly earlier (Wilcoxon $\mathrm{W}=5.76^{*} 10^{\wedge} 10, \mathrm{p}<0.001$) and achieved higher final marks ($\mathrm{W}=5.60^{*} 10^{\wedge} 10, \mathrm{p}<0.001$). The results indicate that behavioral analytics must go beyond simple traffic curves and combine volume, timing, and performance metrics to capture meaningful longitudinal change.
