Abstract
The Mid-infrared ELT Imager and Spectrograph (METIS) is a first-generation instrument for the Extremely Large Telescope (ELT), Europe's next-generation 39 m ground-based telescope for optical and infrared wavelengths, which is currently under construction at the European Southern Observatory (ESO) site at Cerro Armazones in Chile. METIS will offer diffraction-limited imaging, low- and medium-resolution slit spectroscopy, and coronagraphy for high-contrast imaging between 3 and 13 microns, as well as high-resolution integral field spectroscopy between 3 and 5 microns. The main METIS science goals are the detection and characterisation of exoplanets, the investigation of proto-planetary disks, and the formation of planets. The Single-Conjugate Adaptive Optics (SCAO) system corrects atmospheric distortions and is thus essential for diffraction-limited observations with METIS. SCAO will be used for all observing modes, with high-contrast imaging imposing the most demanding requirements on its performance. The Final Design Review (FDR) of METIS took place in the fall of 2022; the development of the instrument, including its SCAO system, has since entered the Manufacturing, Assembly, Integration and Testing (MAIT) phase. Numerous challenging aspects of an ELT Adaptive Optics (AO) system are addressed in the mature designs for the SCAO control system and the SCAO hardware module: the complex interaction with the telescope entities that participate in the AO control, wavefront reconstruction with a fragmented and moving pupil, secondary control tasks to deal with differential image motion, non-common path aberrations and mis-registration. A K-band pyramid wavefront sensor and a GPU-based Real-Time Computer (RTC), tailored to the needs of METIS at the ELT, are core components. This current paper serves as a natural sequel to our previous work presented in Hippler et al. (2018). It reflects all the updates that were implemented between the Preliminary Design Review (PDR) and FDR, and includes updated performance estimations in terms of several key performance indicators, including achieved contrast curves. We outline all important design decisions that were taken, and present the major challenges we faced and the main analyses carried out to arrive at these decisions and eventually the final design. We also elaborate on our testing and verification strategy, and, last not least, comprehensively present the full design, hardware and software in this paper to provide a single source of reference which will remain valid at least until commissioning.

Abstract
The advancement of the use of Artificial Intelligence (AI) in the healthcare sector makes it possible to use computational intelligence applications to assist healthcare professionals in the diagnosis process, facilitating and optimizing early detection and allowing for a more accurate diagnosis (He et al., 2019). The application of machine learning methods, and, more recently, deep learning, has shown promising results (Barbosa et al., 2022; da Silva et al., 2021; De Oliveira et al., 2020; Espinola et al., 2021a, b; Gomes et al., 2021, 2023; Santana et al., 2018; Torcate et al., 2022). These approaches allow powerful tools to support diagnostic imaging and signs to be built, through the extraction of image features and the creation of a classification system, for example (Yu et al., 2018). There are several diseases known and classified by man, with different causes and prevalence. Therefore, contributing to the early detection of diseases defined as neglected was the initial motivation for this work.

Abstract
Most forecasting methods use recent past observations (lags) to model the future values of univariate time series. Selecting an adequate number of lags is important for training accurate forecasting models. Several approaches and heuristics have been devised to solve this task. However, there is no consensus about what the best approach is. Besides, lag selection procedures have been developed based on local models and classical forecasting techniques such as ARIMA. We bridge this gap in the literature by carrying out an extensive empirical analysis of different lag selection methods. We focus on deep learning methods trained in a global approach, i.e., on datasets comprising multiple univariate time series. Specifically, we use NHITS, a recently proposed architecture that has shown competitive forecasting performance. The experiments were carried out using three benchmark databases that contain a total of 2411 univariate time series. The results indicate that the lag size is a relevant parameter for accurate forecasts. In particular, excessively small or excessively large lag sizes have a considerable negative impact on forecasting performance. Cross-validation approaches show the best performance for lag selection, but this performance is comparable with simple heuristics.

Abstract
We present LAZYFS, a new fault injection tool that simplifies the debugging and reproduction of complex data durability bugs experienced by databases, key-value stores, and other data-centric systems in crashes. Our tool simulates persistence properties of POSIX file systems (e.g., operations ordering and atomicity) and enables users to inject lost and torn write faults with a precise and controlled approach. Further, it provides profiling information about the system's operations flow and persisted data, enabling users to better understand the root cause of errors. We use LAZYFS to study seven important systems: PostgreSQL, etcd, Zookeeper, Redis, LevelDB, PebblesDB, and Lightning Network. Our fault injection campaign shows that LAZYFS automates and facilitates the reproduction of five known bug reports containing manual and complex reproducibility steps. Further, it aids in understanding and reproducing seven ambiguous bugs reported by users. Finally, LAZYFS is used to find eight new bugs, which lead to data loss, corruption, and unavailability.

Abstract
Cyberattacks exploit deceptions involving the Domain Name Service (DNS) to direct users to fake websites, such as typosquatting attacks, which exploit natural typographical errors, and homograph attacks, where different Unicode characters resemble the legitimate ones. The deception attacks may also exploit the confusion between DNS domain names, specifically Top-Level Domains (TLDs), and file extensions. Recently, two new TLDs were added, zip and mov, sharing names with certain file types. This overlapping can be explored by malicious actors in a range of threat scenarios to compromise user security. This paper provides an overview of threats originating from the confusion between specific TLDs and file extensions, such as the recent zip and mov. The threats are grouped into 6 threat scenarios that are described and discussed. This research can be part of a more comprehensive strategy that includes addressing the risks associated with these threats and designing future strategies to address the threats associated with exploiting this ambiguity.

Abstract
In recent years, the energy landscape has undergone significant transformations, characterized by the integration of renewable energy sources, smart grids, and the proliferation of IoT-enabled devices. As a result, the efficient management of energy resources has become paramount, requiring advanced methodologies in load forecasting and clustering. This article presents an enhanced methodology for short-term load forecasting that focuses on load consumption profile recognition within a smart building environment. The methodology is designed to analyze and identify recurring load consumption profiles and measures of sensors, thereby enhancing load consumption profile recognition capabilities within the smart building context. The interaction between single and grouped datasets is explored to enhance the accuracy and interpretability of predictions, contributing to optimized energy consumption and providing valuable information for demand response programs. The default forecasting methods used in the methodology are artificial neural networks and k-nearest neighbors. For comparing results and evaluating the proposed approach, XGBoost is also employed. The dataset is selected from a specific database, and the clustering method, partitioning type, is applied with k-means. The results, validated with error evaluation models and statistics, reveal the advantages of the proposed approach, especially with three clusters, where the results achieved by the Artificial Neural Network are the best. The clustering process, particularly the partitioning type, demonstrates a strong capability in improving load forecasting in smart buildings and helps understand load consumption patterns and achieve energy savings. Copyright (C) 2024 The Authors. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)