Abstract
The proliferation of explicit material online, particularly pornography, has emerged as a paramount concern in our society. While state-of-the-art pornography detection models already show some promising results, their decision-making processes are often opaque, raising ethical issues. This study focuses on uncovering the decision-making process of such models, specifically fine-tuned convolutional neural networks and transformer architectures. We compare various explainability techniques to illuminate the limitations, potential improvements, and ethical implications of using these algorithms. Results show that models trained on diverse and dynamic datasets tend to have more robustness and generalisability when compared to models trained on static datasets. Additionally, transformer models demonstrate superior performance and generalisation compared to convolutional ones. Furthermore, we implemented a privacy-preserving framework during explanation retrieval, which contributes to developing secure and ethically sound biometric applications. © 2024 IEEE.

Abstract
In the field of autonomous Unmanned Aerial Vehicles (UAVs) landing, conventional approaches fall short in delivering not only the required precision but also the resilience against environmental disturbances. Yet, learning-based algorithms can offer promising solutions by leveraging their ability to learn the intelligent behaviour from data. On one hand, this paper introduces a novel multimodal transformer-based Deep Learning detector, that can provide reliable positioning for precise autonomous landing. It surpasses standard approaches by addressing individual sensor limitations, achieving high reliability even in diverse weather and sensor failure conditions. It was rigorously validated across varying environments, achieving optimal true positive rates and average precisions of up to 90%. On the other hand, it is proposed a Reinforcement Learning (RL) decision-making model, based on a Deep Q-Network (DQN) rationale. Initially trained in simulation, its adaptive behaviour is successfully transferred and validated in a real outdoor scenario. Furthermore, this approach demonstrates rapid inference times of approximately 5ms, validating its applicability on edge devices.

Abstract
The number of electric vehicles is increasing progressively for various reasons, including economic and environmental factors. There has also been a technological development regarding both the operation and charging of these vehicles. Therefore, it is very important to reinforce the charging infrastructure, which can be optimised through the application of computational tools. There are several approaches that should be considered when trying to find the best location for electric vehicles charging stations. In the literature, different methods are described that can be applied to address this specific issue, including optimisation methods and decision-making techniques such as multicriteria analysis. One of the possible limitations of these methods is that they may not consider all perspectives of the various entities involved, potentially resulting in solutions that do not fully represent the optimal outcome; nevertheless, they provide invaluable information that can be applied in the development of integrative models and potentially more comprehensive ones. This article presents a research and discussion on the most commonly used decision models for this issue, considering optimisation models and multi-criteria decision-making strategies for the adequate planning of EV charging station installation,taking into account the different perspectives of the involved entities.

Abstract
Abstract Breakthroughs in sequencing technologies led to an exponential growth of genomic data, providing unprecedented biological in-sights and new therapeutic applications. However, analyzing such large amounts of sensitive data raises key concerns regarding data privacy, specifically when the information is outsourced to third-party infrastructures for data storage and processing ( e . g ., cloud computing). Current solutions for data privacy protection resort to centralized designs or cryptographic primitives that impose considerable computational overheads, limiting their applicability to large-scale genomic analysis. We introduce G yosa , a secure and privacy-preserving distributed genomic analysis solution. Unlike in previous work, G yosa follows a distributed processing design that enables handling larger amounts of genomic data in a scalable and efficient fashion. Further, by leveraging trusted execution environments (TEEs), namely Intel SGX, G yosa allows users to confidentially delegate their GWAS analysis to untrusted third-party infrastructures. To overcome the memory limitations of SGX, we implement a computation partitioning scheme within G yosa . This scheme reduces the number of operations done inside the TEEs while safeguarding the users’ genomic data privacy. By integrating this security scheme in Glow , G yosa provides a secure and distributed environment that facilitates diverse GWAS studies. The experimental evaluation validates the applicability and scalability of G yosa , reinforcing its ability to provide enhanced security guarantees. Further, the results show that, by distributing GWASes computations, one can achieve a practical and usable privacy-preserving solution.

Abstract
This study introduces a novel Demand Response (DR) Model based on Multi-agent Reinforcement Learning (MARL-DR), comprised of a pricing and incentives scheme aimed at improving the accuracy of existing demand response strategies. Furthermore, this new approach represents a flexibility solution to prevent sharp price variations caused by the high penetration of unconventional renewable energy, which are directly passed on to end-users. The model employs a cooperative-competitive MARL-DR technique based on Q-learning, with the goal of determining optimal prices and incentives that maximize benefits for both customers and electric Service Provider (SP). In this regard, the model has the capability to offer customers pricing options in both real-time and time-of-use, to actively adjust each user's consumption. Additionally, through demand characterization factors, such as the coincidence factor (CF), the of typical user behavior is improved, and the influence of individual user demand on system peak demand is more accurately detected. It is also demonstrated that the cooperative-competitive approach offers better performance compared to other approaches. Finally, a sensitivity analysis is presented at various stages of the model to verify its accuracy and efficiency in pricing and incentives formulation.

Abstract
Offshore wind farms face harsh maritime conditions, prompting the use of sacrificial anodes to prevent rapid structural degradation. Regular maintenance and replacement of these elements are vital to ensure ongoing corrosion protection, maintain structural integrity, and optimize efficiency. This article details the design and validation of the MARESye hybrid underwater imaging system, capable of retrieving heterogeneous tri-dimensional information with millimetric precision for the close-range inspection of submerged critical structures. The optical prowess of the system is first validated during low turbidity trials where the volumetric properties of a decommissioned anode are reconstructed with absolute errors down to 0.0008 m, and its spatial dimensions are depicted with sub-millimeter precision accounting for relative errors as low as 0.31%. MARESye is later equipped as payload in a commercial ROV during areal environment inspection mission at the ATLANTIS Coastal Test Center. This experiment sees the sensor provide live reconstructions of a sacrificial anode, revealing a biofouling layer of approximately 0.0130 m thickness. The assessment of the high-fidelity 2D/3D information obtained from the MARESye sensor demonstrates its potential to enhance the situational awareness of underwater vehicles, fostering reliable O&M procedures.