Vol. 41 · No. 1

Given the importance of the computer systems in our daily life today, it is decisive to be able to protect the computer systems against attacks. Intrusion Detection Systems (IDSs) are the crucial component of modern cybersecurity systems. IDSs are built-in in the devices of the major providers such as Cisco and Juniper. Since the early days of the Internet up to now, the IDSs rely heavily on signature-based detection methods. However, in recent years, researchers utilize the power of machine learning techniques and achieve very good performance in classifying network attacks. In this paper, we analyze the machine learning techniques that have been proposed in recent years. We propose some new techniques to improve the performance of the existing methods. The experimental results using real-world datasets show that our suggestions can boost the predictive accuracy of the models.

Cloud-based manufacturing is taking shape, and many industries seem interested to make the transition to it. Developing blockchain solutions for trusted computing is also taking its roots. Developing a blockchain-based solution for cloud-based manufacturing systems is a field that is new but also faces limitations and a lack of case studies. Smart contracts are one part of the solution which deals with making blockchain successful in cloud-based manufacturing. As we move towards smart contracts design and development for cloud-based manufacturing, there is no complete survey of smart contract and cloud manufacturing that can highlight critical, challenging issues and limitations. Most of the work found in smart contracts is mostly financial and notary-centric applications. On the cloud manufacturing side, most of the literature deals with Internet of Things (IoT) and cloud computing systems. Therefore, there is a need to study the best practices to start manufacturing supported by blockchain smart contracts. We conducted a scoping review for smart contracts for cloud manufacturing to address the problem mentioned above. We studied the latest case studies and concepts in data extracted from digital libraries and online repositories. Furthermore, we follow the relevance and acceptance criteria of research articles for inclusion and exclusion from this work. This paper focuses on blockchain systems, smart contracts and architecture, smart contracts in the cloud, and the IoT environment. Furthermore, we tried to bridge design and implementation details for readers to understand the patterns that can replicate for cloud-based manufacturing systems.

The number of Twitter users is increasing and the quantity of produced data is growing. Using this big data to analyze user behavior has become a very active field. The two key challenges of this paper are extracting data from Twitter and extracting topics from user tweets. The proposed approach uses data crawling to collect data from Twitter and a bunch of natural language processing techniques to extract information from the so collected data and build a dataset. Thereafter, we use K-means clustering and Latent Dirichlet Allocation to extract the prevalent topics from this dataset, as they are the most common in the literature. Our proposal is generic, it can be reused by scientists to annotate any text collection.

Given a text, can we segment it into semantically coherent sections in an automatic way? Can we detect the semantic boundaries, if we know how many they are? Can we determine how many semantically distinct sections are in the text? These are the questions we address in this paper. To respond, we use the Bidirectional Encoder Representation from Transformer (BERT) to analyze the text and evaluate a function that we call local incoherence, which we expect to show maxima at the points where a semantic boundary is detected. Our results, although preliminary, are encouraging and suggest that our approach can be successfully applied. However, they are quite sensitive with respect to the text quality, as it happens in the case in which the text is derived from an audio stream via Automatic Speech Recognition techniques.

Due to the recent escalation in the amount of text data available and used online, text classification has become a staple for data analysts when extracting relevant information. Yet, machine learning algorithms are susceptible to biases when implemented on any large-scale automated task, especially in text analysis. With the popularization of newer branches of study emerging from the field of machine learning – such as ensemble and deep learning – we must analyze the potential pitfalls in the common experimental setup centered around learning algorithms. Imbalance in text data is one such pitfall – when data is not equally distributed across all categories in a dataset, it can influence and undermine the classification of underrepresented categories. In our research, we have proposed several techniques and unique approaches to tackle this obstacle. We prepared four datasets of varying degrees of imbalance to conduct our experimentation. We proved that feature extraction techniques singular value decomposition (SVD) and GloVe are the key to reducing the effect of imbalance in text classification, especially in ensemble and deep learning. Using the result of our research, we have also proposed a modified ensemble classifier that can classify imbalanced and balanced data alike.

Active systems (ASs) are a special class of (asynchronous) discrete-event systems (DESs). An AS is represented by a network of components, where each component is modeled as a communicating automaton. Diagnosing a DES amounts to finding out possible faults based on the DES model and a sequence of observations gathered while the DES is being operated. This is why the diagnosis engine needs to know what is observable in the behavior of the DES and what is not. The notion of observability serves this purpose. In the literature, defining the observability of a DES boils down to qualifying the state transitions of components either as observable or unobservable, where each observable transition manifests itself as an observation. Still, looking at the way humans observe reality, typically by associating a collection of events with a single, abstract perception, the state-of-the-art notion of DES observability appears somewhat narrow. This paper presents, a generalized notion of observability, where an observation is abstract rather than concrete, since it is associated with a DES behavioral scenario rather than a single component transition. To support the online diagnosis engine, knowledge compilation is performed offline. The outcome is a set of data structures, called watchers, which allow for the tracking of abstract observations.

Providing and distributing fresh water to large communities is a major global concern. In addition to its scarcity as well as to its wastage, this vital resource is being affected by challenging environmental conditions. New approaches are, therefore, urgently needed for an optimized, fair, and efficient use of fresh water. The adoption of emergent technologies is giving high hopes to reach this objective. Among these technologies, digital twin is attracting increasing attention from the academic and industrial committees. This attention is particularly motivated by its expected values to any sector, including process optimization, cost reduction, and time to market shortening. In the specific field of water management, several solutions are being proposed, especially to detect leak detection and test water assets under a variety of working constraints. These solutions are still lacking intelligence and autonomy throughout the loop of data acquisition and processing as well as asset control and service generation and delivery. To this end, we are proposing in this paper a new framework based on multi-agent systems and DT paradigm to close gaps within this loop. Our multi-agent system is responsible of running data analytics mechanisms in order to assess water consumption and generate relevant feedbacks to users using, among others, a rewarding system to select the appropriate pricing policies. It is also responsible of simulating asset operations under specific working constraints for the purpose of failure and/or defect detection.

This research lays down foundations for a stronger presence of machine learning in the emergency department. Using machine learning to make predictions on a patient's situation can increase patient's health and decrease the waiting time. This paper explores to what extent it is possible to accurately predict ER outcome. These predictions will be based on routinely available ER data from a Dutch hospital. The data set used is representative for any Dutch Hospital. Prediction performance is compared between ML predictors. Using random forest and stacked ensemble gathered the best results. This research found that for more than half of the adult patients, the algorithm can very accurately predict hospitalization, with similar results for children and during the COVID-19. Moreover, it is investigated which characteristics and events contribute to the direction of the patient. Finally, several plans are introduced to substantially improve the ER process, for example by quickly reviewing patients selected by the algorithms. These might lead to an ER process that is significantly quicker, with more accurate diagnosis.

Recently, combinatorial optimization problems have aroused a great deal of interest in Machine Learning, leading to interesting advances in Neural Combinatorial Optimization (NCO): the study of data-driven solvers for NP-Hard problems based on neural networks. This paper studies the benefit of Transfer Learning for NCO by evaluating how model training can be improved taking advantage of knowledge learned while solving similar tasks. We focus, in particular, on two famous routing problems: the Traveling Salesman Problem (TSP) and the Vehicle Routing Problem (VRP). The latter being a generalization of the former, we study the effect of applying Transfer Learning from a model trained to solve TSP while training a model learning to solve the Capacitated VRP (CVRP). We present adaptations of a state-of-the-art NCO model for implementing Transfer Learning. Our results based on extensive empirical experiments in different settings show that Transfer Learning may help to speed up the training process while being more sample efficient.

The healthcare sector is a collaborative environment that requires a joint action for delivering care. Health professionals who work in different organizations or settings must assimilate a massive amount of data generated during the patient care journey. Electronic healthcare records offer a starting point for supporting cooperation among healthcare professionals by saving and sharing traces of the patient's medical acts. However, we claim that these records merely store and share data, which disregards how health professionals use this data to understand the patients' situations and make decisions. We argue that focusing on the cooperative practices of managing patients gives designers new insights to design future healthcare information systems supporting cooperation, and we identify challenges related to this design approach.

Hospitals are facing a wide variety of challenges in terms of quality and efficiency of healthcare. Internet of Things (IoT) is a technology used by organisations to increase efficiency and quality by recording measurements for historic analysis. The data thus produced can then go on to inform future decisions and predictions. Unfortunately, the benefits provided for by a successful IoT adoption are currently out of reach for many hospitals. The lack of a maturity model for IoT adoption in hospitals aggravates this situation. The goal of defining and applying such a model is to assist hospitals in reaching a higher level of IoT maturity and thereby improving the quality of services delivered. This paper develops an IoT maturity model that is tailored to the healthcare industry with an emphasis on Belgian hospitals. The developed maturity model is grounded in scientific literature and industry expert opinions. Experts found the maturity model to be relevant, clear, and helpful for the hospitals' road to IoT adoption.

The lack of electronic health record data in general and especially at initial phases of medical research projects is common and is one of the main reasons for delay or failure of such projects. One of the health areas with little attention is the home care area, where patients are being supported by their families or informal caregiver at home. In this paper we present related work on medical data formats and synthetical data generation of medical health records. Furthermore, it presents an approach to generate synthetic electronic health records (HER) that are readily available; suited to research; and free of legal, privacy, security and intellectual property restrictions to be used in home care research projects. We adapted and used Synthea™, an open-source software framework that simulates the lifespans of synthetic patients to generate synthetic EHRs. This paper presents the use case of home care from the capturing of user requirements of home care patients, translating the requirements into a data model, feeding the data model into Synthea™ framework, which produces synthetical health data records mainly as QuestionnaireResponse instance of the Fast Healthcare Interoperability Resources (FHIR) to using these EHRs to build an initial machine learning data model for home care.

The identification of tumorous tissues in the brain based on Magnetic Resonance Images (MRI) analysis is a challenging and time consuming task that highly depends on radiologists expertise. As prompt diagnosis of tumors can often be inherent to the patient's survival, it is however crucial to decrease the amount of time spent on the manual analysis of MRI while increasing the accuracy of the detection process. To tackle these issues, many research works have already investigated efficient computer vision systems. They offer new opportunities to assist health care providers in the establishment of fast and more accurate tumor detection, classification and segmentation. However, often based on deep learning methods, the development and tuning of these solutions remains time and energy consuming while inducing a lack of explainability in the decision making system. In this study, we respond to these issues by solving a brain tumor detection task using the Selective Search (SS) algorithm coupled with a simplified Pulse-Coupled Neural Network (PCNN) for visual feature extraction and detection validation. The performed experiments showed promising results in terms of computational cost and detection accuracy. This leads to the development of a light-weight brain tumor detection system.

Industrial companies are more and more interested in the use of artificial intelligence (AI) in the control and monitoring of their processes. They try to take advantage of the power of this technology in order to increase the level of automation and to build smarter machines with new capabilities of self-adaptation and self-control. Especially, the automotive industry, with their high requirements in productivity and diversity management, are eager to adapt AI concepts to their processes. However, the training of Deep Learning (DL) models requires an important effort of data preparation, providing a dataset of all possible configurations. Indeed, this dataset must be collected and then annotated. Considering the fact that automotive industry deals with a huge number of references and that it often and quickly needs to modify their products, it is very difficult, if not impossible, to gather sufficient datasets for each produced reference and to have the time to train DL models in the plants with the traditional methods. This paper presents an innovative methodology to prepare the dataset by creating virtual images instead of collecting real ones and then automatically annotating them. It will demonstrate that this method will reduce the efforts and the time of the preparation of the dataset significantly. The paper will also present how this method was deployed for the quality control of welding operations in the automotive industry.

This paper deals with the home health care service (HHCS) which is defined as a set of medical, paramedical and social services delivered to patients in their domicile rather than in hospital. To support decision making in HHC, optimization models have been used. However, several of those models are deterministic and do not address the dynamical and uncertainty aspects of the system and variability of some patient data. The HHC scheduling problems are facing more and more complex and specific constraints. These constraints have to be respected, meanwhile the problem objective is optimized under parameters uncertainties. This paper aims to formulate a model that integrated home care scheduling problem while taking into account human aspect – the behavior of patients, and also another aspect like travel time uncertainty and dynamic behavior of involved medical team and social actors in HHC. Robust approaches are adopted to model and handle this uncertainty.

eHealth is widely recognized as the application of information and communication technologies (ICT) in health. However, eHealth initiatives are still incipient in Latin America and the Caribbean, and, in many cases, restrictions in terms of access have been reported, also the necessary infrastructure, interoperability and scalability of these. This work proposes a set of critical success factors (CSFs) for the implementation of eHealth, which allow the identification of gaps, and the proposal of alternatives for the optimization of eHealth. It starts with the establishment of an eHealth domain, its scope and contributions, prior to the identification of key topics, and the establishment of every CSF, with support in guiding questions and metrics. The CSFs can facilitate the planning of projects or activities in eHealth, promoting strengths, either in ICT, management, or another of the involved topics. The CSFs must be employed with a criterion of flexibility and adequacy regarding the case in which they are applied. Finally, opportunities to evaluate and apply the CSFs in a specific context are set out.

The Internet of Things (IoT), as a concept, was not officially named until 1999 where it was still used by big computer and communication companies. It is the connection between objects anywhere, anytime, using internet communication. IoT is one of the network concepts which are growing rapidly in the last few years. The connected devices reach billions which leads to a huge increase in data transfer through the network. This rapid increase of transferred data is overloading network servers which result in more processing and routing time. Fog computing and cloud computing paradigms extend the edge of the network, thus enabling a new variety of applications and services. In this research, we focus on the processing and routing time, moreover, we present a new model in the application layer of the IoT system to classify IoT applications according to their valued data. Also, we work on modeling the fog computing architecture and use the cell operator as the main fog center to store data and compare its performance with the traditional model. We present a comparative study with the traditional IoT architecture based on classifying applications and define a priority for each application. We aim to give solutions to lower data transmission time, reduce routing processes, decrease internet usages, increase response speed, deliver important and sensitive data first, improve the quality of services, enhance the overall performance of IoT systems by depending on fog network as the main layer for processing and storing data, then by giving each application a priority value to be served according to it where the application with the highest priority is served first on the network. Our method which is based on static priority shows better performance and management against the RWS and DRAG method which are based on many parameters to take a decision.

Enterprise Application Integration (EAI) and HL7 (Health Level Seven) messaging are well established technologies in healthcare environments. Due to the age and longevity of HL7 standards (especially HL7 V2.x) and their widespread use, many interfaces outlive the middleware on which they run and must be ported to new systems. This often requires the entire code of the interface to be rewritten, which is associated with great effort and costs. This paper shows a generic EAI framework based on configuration and dependency injection for implementing reusable interfaces upfront and the results when applied to a real production EAI environment of an Austrian healthcare provider.

Road traffic is drastically increasing in big cities around the world. In order to enable a flexible management of this traffic, Intelligent Transportation System (ITS) solutions are relying on emergent ubiquitous, mobile, and communication technologies, particularly to intelligently deal with the limited capacities of the existing road infrastructures. While intelligence is left to the autonomous and connected vehicles as well as to the ITS, the road infrastructure has been mostly playing a passive role (as a source of data). Road signage, in particular, are in best cases dynamic but do not play an active role in monitoring traffic and incidents. We propose in this paper to build Smart Road Signs (SRS) that can collaborate with Connected Vehicles in order to monitor traffic and warn drivers about any incident or danger. Our SRSs are meant to operate autonomously in order to detect road traffic problems, share appropriate information with vehicles in the vicinity, and display relevant messages based on the ongoing contextual situation. To meet our goals, we rely on Multi-Agent Systems to design SRSs as proactive components in the ITS landscape. We also rely on agent mobility in order to strengthen the collaboration with the connected vehicles.