Multitask learning (MTL) has emerged as a promising paradigm for improving recommendation systems by learning multiple related tasks together. In this paper, we present significant improvements in personalized news recommendation by integrating auxiliary and cascaded tasks. Our study compares single-task learning (STL) models with multitask architectures. We evaluated performance on the primary task of news-click prediction, along with predicting user interest in news categories and topics as auxiliary tasks, and fully scrolled prediction and reading pattern prediction as cascaded tasks. Our results indicate that MTL models outperform STL baselines in terms of AUC metrics. These results underscore the benefits of using multiple related tasks to capture richer signals of user behavior, while also highlighting challenges that remain, such as effectively integrating non-click samples in cascaded tasks.

The increasing convergence of Operational Technology (OT) networks into Information Technology (IT) communications poses critical infrastructures to new threats that may cause huge hazards. The study of protection mechanisms and the development of security systems capable of preventing such attacks is of paramount importance nowadays. Besides formally defining the model representing the intertwining of IT and OT networks of a Chemical Industry, we prove the ability to detect different types of attacks with good results experimentally by implementing an Intrusion Detection System (IDS) based on Deep Learning (DL) that achieves an accuracy of 87, 19%.

The increasing number and variety of cyber attacks in recent years have made intrusion detection systems (IDS) a critical component of computer network defense to monitor network traffic and identify malicious activities. Machine learning (ML) and deep learning (DL) techniques have been increasingly used in anomaly-based network IDS (NIDS) to detect new and unknown attacks, but they have been proven to be vulnerable to adversarial attacks, which can significantly reduce the detection system performance. In this paper we investigate the robustness of a DNNs-based NIDS, implemented for the Secure Safe Apulia Progect, against adversarial untargeted white box attacks. We employ Fast Gradient Sign Method (FGSM) and Projected Gradient Descent (PGD) as adversarial attacks to evaluate the decrease in model accuracy. The results show that adversarial training is an effective defense strategy against these types of attacks allowing the model to achieve F1 score values of 93%, 99%, 85%, 83% respectively, for the classification of benign instances, Backdooor, Ransomware and XSS malicous instances. This work aims to contribute to the challenge of handling adversary attacks in the domain of NIDS, in which research is still moving its first steps.

In clinical practice, patient care flows are generally subject to recommended and standardized therapeutic interventions. Especially in a home care setting, situation-aware adherence to therapy can be both difficult for the patient to follow and difficult for the physician to assess. Process mining techniques may be useful artificial intelligence solutions for remotely assessing the compliance of patients' behavior with the corresponding care path, especially if adopted in a cognitive IoT Edge infrastructure, dedicated to the acquisition and analysis of daily routines in a form of event log. In this paper, we present an innovative method to measure in-home adherence to metabolic syndrome management with the aim of providing awareness of the patient's current situation. The analytical results demonstrate the validity of using process mining techniques to remotely evaluate patient behavior.

Filter face masks are Respiratory Protective Equipment designed to protect the wearer from various hazards, suit various health situations, and match the specific requirements of the wearer. Current traditional face masks have several limitations. In this paper, we present (ME)(2), the Monitoring Equipment Mask Environment: an innovative reusable 3D-printed eco-sustainable mask with an interchangeable filter. (ME)(2) is equipped with multiple vital sensors on board, connected to a system-on-a-chip micro-controller with computational capabilities, Bluetooth communication, and a rechargeable battery that allows continuous monitoring of the wearer's vital signs. It monitors body temperature, heart rate, and oxygen saturation in a non-invasive, strategically positioned way. (ME)(2) is accompanied by a mobile application that provides users' health information. Furthermore, through Edge Computing Artificial Intelligence (Edge AI) modules, it is possible to detect an abnormal and early symptoms linked to possible pathologies, possibly linked to the respiratory or cardiovascular tract, and therefore perform predictive analysis, launch alerts, and recommendations. To validate the feasibility of embedded in-app Edge AI modules, we tested a machine learning model able to distinguish COVID-19 versus seasonal influenza using only vital signs. By generating new synthetic data, we confirm the highly reliable performances of such a model, with an accuracy of 94.80%.

The progress of the Internet of Medical Things (IoMT) and mobile technologies is a crucial driver for the evolution of healthcare systems in the path of prevention, early diagnosis and care, consequently unleashing the full potential of medical devices. Especially in intensive care, several vital signs can be monitored to provide an Early Warning Score (EWS) useful to detect the onset of pathological events or severe conditions. However, under these conditions, it would be beneficial to design a system that can provide a risk assessment even in the presence of a reduced number of vital signs. In this work, we propose an on-edge system, connected to one or more wearable medical devices, that is able to collect, analyze and interpret real-time clinical parameters and to provide an EWS-like clinical risk measurement. The system shows an evolutionary behavior by dividing the learning problem in two simpler ones, in order to correctly distinguish between low-urgency and emergency scenarios, with the possibility of selecting the most convenient configuration able to choose the most appropriate classifier even when the feature set does not allow a robust model selection. In particular, we focus on a comparative analysis of machine learning (ML) methods in different conditions of available vital parameter sets, evolving therefore to an adaptive ML approach. Moreover, since from the integration of artificial intelligence tools and IoMT, emerging ethical issues may arise about lack of transparency, a semantic-based explanation is associated to enrich the predictions along with the health data by means of Semantic Web technologies.

Nowadays citizens engage with smart city ecosystems in several ways using smartphones, mobile devices, connected cars, and drones. Pairing devices and data with a city's infrastructure and services can improve sustainability and achieve an improvement in awareness and territorial control. Communities can improve energy distribution and decrease traffic congestion with the help of IoT technologies. To support and streamline such a process, in this paper we introduce a Web crowdsourcing platform as a Common Operational Picture dashboard to interoperate with smart devices, collect urban data from them, and monitor the city in real-time. Its application to the Metropolitan City of Bari is presented and discussed.

A Virtual Private Network (VPN) ensures the confidentiality and integrity of data transferred between two endpoints, even if the means of transport are insecure. One popular protocol is Internet Protocol Security (IPSec) which operates at OSI layer 3 and protects all protocols at higher layers. Cryptographic keys in IPSec are negotiated using the Internet Key Exchange (IKE) protocol. IKE negotiates security parameters for IPSec sessions. In particular, the IKEv2 protocol uses the Elliptic Curve Diffie-Hellman (ECDH) algorithm to establish a secret key shared between two nodes on a network. Although solving such a problem remains difficult with current computing power, it is believed that generic quantum computers will be able to solve this problem, which implies that the security of IKEv2 is compromised. There are, however, several cryptographic systems that are trusted to be resistant to attacks by quantum computers. This family of cryptosystems is known as quantum-resistant cryptography (QRC). In this paper, after highlighting the requirements for a secure key exchange protocol, we briefly review the QRC solutions that have been proposed in the recent literature.

Nowadays, modern technology is widespread in sports; therefore, finding an excellent approach to extracting knowledge from data is necessary. Machine Learning (ML) algorithms can be beneficial in biomechanical data management because they can handle a large amount of data. A fencing lunge represents an exciting scenario since it necessitates neuromuscular coordination, strength, and proper execution to succeed in a competition. However, to investigate and analyze a sports movement, it is necessary to understand its nature and goal and to identify the factors that affect its performance. The present work aims to define the best model to screen élite and novice fencers to develop further a tool to support athletes’ and trainers’ activity. We conducted a cross-sectional study in a fencing club to collect anthropometric and biomechanical data from élite and novice fencers. Wearable sensors were used to collect biomechanical data, including a wireless inertial system and four surface electromyographic (sEMG) probes. Four different ML algorithms were trained for each dataset, and the most accurate was further trained with hyperparameter tuning. The best Machine Learning algorithm was Multilayer Perceptron (MLP), which had 96.0% accuracy and 90% precision, recall, and F1-score when predicting class novice (0); and 93% precision, recall, and F1-score when predicting class élite (1). Interestingly, the MLP model has a slightly higher capacity to recognize élite fencers than novices; this is important to determine which training planning and execution are the best to achieve good performances.

The adoption of digital tools in the health sector is a need driven by optimizing processes and improving collaboration between different stakeholders. Exploiting data is the principal activity of the eHealth sector, even vital for some patients' clinical conditions because its monitoring can help prevent adverse events or disease degeneration. However, the actual use of electronic clinical information to evaluate human decisions against predefined protocols or statistically known evolution patterns is still mostly under-exploited. Clinical Pathway is the primary tool for implementing clinical guidelines and evidence-based medicine. It is used to improve the care processes by monitoring changes in clinical practices to reach the best appropriate care more quickly and reduce the health system's costs. In this work, we present an Edge architecture for Ambient Assisted Living in the context of home hospitalization. Using process mining techniques helps understand patients' behaviours, assess their compliance with the corresponding clinical path, and support physicians in making decisions broadly and transparently.

Recent advances in quantum computing and quantum information theory represent a severe threat to the current state of the art of data protection. In this context, new quantum-safe techniques have emerged in recent decades, which fall into post-quantum and unconditionally secure cryptographic schemes. The firsts rely on computational problems supposed to be hard also for quantum computers. In contrast, the seconds do not depend on the difficulty of a computational problem and are therefore immune to quantum power. In particular, unconditionally secure techniques include Quantum Key Distribution (QKD) protocols for transmitting secret keys thanks to the quantum properties of light. In this work, we discuss QKD networks and post-quantum algorithms, considering their opportunities and limitations and showing that reconciliation between these two directions of cryptography is feasible and necessary for the quantum era.This work is part of the activities of the PON project “Development of quantum systems and technologies for IT security in communication networks” (QUANCOM) which aims to the realization of a metropolitan quantum communication network through the collaboration between universities, research centers and companies operating in the communication market area.

Recent advances in neuroimaging techniques, such as diffusion tensor imaging (DTI), represent a crucial resource for structural brain analysis and allow the identification of alterations related to severe neurodegenerative disorders, such as Alzheimer's disease (AD). At the same time, machine-learning-based computational tools for early diagnosis and decision support systems are adopted to uncover hidden patterns in data for phenotype stratification and to identify pathological scenarios. In this landscape, ensemble learning approaches, conceived to simulate human behavior in making decisions, are suitable methods in healthcare prediction tasks, generally improving classification performances. In this work, we propose a novel technique for the automatic discrimination between healthy controls and AD patients, using DTI measures as predicting features and a soft-voting ensemble approach for the classification. We show that this approach, efficiently combining single classifiers trained on specific groups of features, is able to improve classification performances with respect to the comprehensive approach of the concatenation of global features (with an increase of up to 9% on average) and the use of individual groups of features (with a notable enhancement in sensitivity of up to 11%). Ultimately, the feature selection phase in similar classification tasks can take advantage of this kind of strategy, allowing one to exploit the information content of data and at the same time reducing the dimensionality of the feature space, and in turn the computational effort.

In the e-Health domain, new and continuously evolving threats emerge every day. The security of e-Health telemonitoring systems is no longer negligible. In this paper, we propose a Cyberattack Detection System (CADS) model that exploits artificial intelligence techniques to detect anomalies without requiring a security analyst, explain the malicious activity, and display suspected attack data to healthcare personnel for feedback. The system description is contextualized to the case of the hacked remote patient health telemonitoring.

New and evolving threats emerge every day in the e-Health industry. The safety of e-Health’s telemonitoring systems is becoming a prominent task. In this work, starting from a CADS (Cyberattack Detection System) model that uses artificial intelligence techniques to detect anomalies, we focus on the activity of interacting with data. Using a User Interaction Engine, a dashboard allows you to visually explore and view data from suspected attacks on healthcare professionals for a threat reaction. In particular, a User Feedback module is presented to interact with healthcare personnel and ask for a response on the anomaly detected.

The SARS-CoV-2 pandemic has brought unexpected new scenarios in patient-care journeys and has accelerated this innovative process in the healthcare sector, demonstrating the importance of a systemic rethinking of remote care, mostly when patients are discharged from the hospital and continue their therapies at home in autonomy. The possibility to remotely monitor patients at home by means of smart sensors and medical devices has a dramatic impact on the quality of health services. Situation awareness plays an essential role in the decision-making process about the users, patients in this case, and their behaviors. Leveraging an Edge Computing framework, with embedded Artificial Intelligence capabilities to process near real-time data gathered from connected smart devices, would provide automatic decision support, thus improving the physicians’ course of action. In this paper we introduce, within an Edge AI framework, a dedicated module, called Clinical Pathway Adherence Checker (CPAC), which identifies the discrepancies between the modeled clinical pathway and the observed one by means of process mining techniques, and hence detecting early clinical deterioration of patient conditions. Also, further analyses are conducted in the anomaly detection at the Edge that may occur during the health data transmission process.

The possibility of continuous monitoring of health conditions represents a crucial aspect for the improvement of living conditions, the prevention of potential pathologies and prompt response in critical situations. In particular, in intensive care or emergency situations, the evaluation of illness degree of a clinical risk level can be considered a predictive task in situations where streams of vital signs data are gathered by medical devices and the Internet of Medical Things (IoMT) sensors. In this framework, Early Warning Score (EWS) systems generating an aggregate score based on the measurement of a set of vital signs, such as National Early Warning Score 2 (NEWS2) and Modified Early Warning Score (MEWS), may provide a helpful decision support for the estimation of health state and triggers for critical care intervention. In the present work, we address a preliminary analysis in order to investigate the most suitable Machine Learning (ML) technique for the prediction of clinical risk classes of a continuously monitored patient in a particular condition where a limited number of vital parameters is available. This analysis is then intended to be preparatory for the final goal of designing an edge device connected to one or more wearable medical devices via IoMT, which adaptively exploits the best ML model to predict a reliable EWS.

Situation awareness is a renowned approach leading to a decision. The results obtained by measuring the awareness of stakeholders in healthcare can provide valuable inputs on the decision-making process. In the world of an ageing society, artificial intelligence in eHealth plays an essential role in the decision-making process of the users and their behaviors. In particular, clinical pathways, as evidence-based patient-care algorithms, describe the process of care for specific medical conditions within a localized setting. In clinical practice, patient-care journeys are generally subject to the recommended treatment interventions which are regulated in clinical pathways. However, unexpected scenarios occur in patient-care journeys and have a dramatic impact on health service delivery of clinical pathways, often when patients are discharged from the hospital and continue to be followed at home. In order to be able to quickly adapt to arising problems or deviations during the execution of clinical pathways, it is very important to be able to monitor clinical pathways in a near real-time manner so as to obtain a current overview of patient care. For this reason, telemedicine, along with the medical and wearable devices, that can now be employed to gather large amounts of data and perform data modeling through artificial intelligence techniques, may improve the clinical pathway management and reduce costs. Therefore, in this paper we introduce a novel Edge Computing framework that encompasses the different applications of telemedicine, ranging from the modeling and adherence to the clinical pathway to the early discovery of clinical deterioration conditions, allowing diagnosis and/or remote treatment through a set of artificial intelligence tools, also assessing the security and privacy issues that may occur during the health data transmission process, thus yielding a situation awareness for eHealth.

From robots that replace workers to robots that serve as helpful colleagues, the field of robotic automation is experiencing a new trend that represents a huge challenge for component manufacturers. The contribution starts from an innovative vision that sees an ever closer collaboration between Cobot, able to do a specific physical job with precision, the AI world, able to analyze information and support the decision-making process, and the man able to have a strategic vision of the future.

One of the challenges launched by the emergence of the Internet of Things (IoT) phenomenon is the integration and harmonization of data produced by cyber-physical systems with those already present and generated by classical information systems. Things, on the other hand, are items that have physical sensors, devices actuators, or other tools to generate data and receive information from other sources. In this complex scenario, we can already talk about the Internet of Everything (IoE), an immense network of heterogeneous and complex elements. In this paper, we present a novel business solution supported by the Italian POR FESR 2014 - 2020 of Apulia Region, namely the Digital Future project, that acts as a high-level framework of applications for real-time platforms, integrating a wide variety of IoT services, leveraging data into more useful information, into interoperable industrial control networks to put humans in the loop, connecting them in a more relevant, valuable ways.