Abstract
Information systems depend on security mechanisms to detect and respond to cyber-attacks. One of the most frequent attacks is the Distributed Denial of Service (DDoS): it impairs the performance of systems and, in the worst case, leads to prolonged periods of downtime that prevent business processes from running normally. To detect this attack, several supervised Machine Learning (ML) algorithms have been developed and companies use them to protect their servers. A key stage in these algorithms is feature pre-processing, in which, input data features are assessed and selected to obtain the best results in the subsequent stages that are required to implement supervised ML algorithms. In this article, an innovative approach for feature selection is proposed: the use of Visibility Graphs (VGs) to select features for supervised machine learning algorithms used to detect distributed DoS attacks. The results show that VG can be quickly implemented and can compete with other methods to select ML features, as they require low computational resources and they offer satisfactory results, at least in our example based on the early detection of distributed DoS. The size of the processed data appears as the main implementation constraint for this novel feature selection method.

Abstract
The government and industry have given the recent development of the Internet of Things in the healthcare sector significant respect. Health service providers retain data gathered from many sources and are useful for patient diagnostics and research for pivotal analysis. However, sensitive personal information about a person is contained in healthcare data, which must be protected. Individual privacy protection is a crucial concern for both people and organizations, particularly when those firms must send user data to data centers due to data mining. This article investigated two general states of increasing entropy by changing the entropy of the class set of characteristics based on artificial intelligence and the k-anonymity model in privacy in context, and also three different strategies have been investigated, i.e., the strategy of selecting the feature with the lowest number of distinct values, selecting the feature with the lowest entropy, and selecting the feature with the highest entropy. For future tasks, we can find an optimal strategy that can help us to achieve optimal entropy in the least possible repetition. The results of our work have been compared by lightweight and MH-Internet of Things, FRUIT methods and shown that the proposed method has high efficiency in entropy criteria.

Abstract
Sensing will be a key technology in 6G networks, enabling a plethora of new sensing-enabled use cases. Some of the use cases require deployments over a wide physical area that needs to be sensed by multiple sensing sources at different locations. The efficient management of the sensing resources is pivotal for sustainable sensing-enabled mobile network designs. In this paper, we provide an example of such use case, and argue the energy consumption due to sensing has potential to scale to prohibitive levels. We then propose architectural enhancements to solve this problem, and discuss energy saving and energy efficient strategies in sensing, that can only be properly quantified and applied with the proposed architectural enhancements.

Abstract
[No abstract available]

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
Face recognition applications have grown in parallel with the size of datasets, complexity of deep learning models and computational power. However, while deep learning models evolve to become more capable and computational power keeps increasing, the datasets available are being retracted and removed from public access. Privacy and ethical concerns are relevant topics within these domains. Through generative artificial intelligence, researchers have put efforts into the development of completely synthetic datasets that can be used to train face recognition systems. Nonetheless, the recent advances have not been sufficient to achieve performance comparable to the state-of-the-art models trained on real data. To study the drift between the performance of models trained on real and synthetic datasets, we leverage a massive attribute classifier (MAC) to create annotations for four datasets: two real and two synthetic. From these annotations, we conduct studies on the distribution of each attribute within all four datasets. Additionally, we further inspect the differences between real and synthetic datasets on the attribute set. When comparing through the Kullback-Leibler divergence we have found differences between real and synthetic samples. Interestingly enough, we have verified that while real samples suffice to explain the synthetic distribution, the opposite could not be further from being true.