O avanço da Internet das Coisas Médicas (IoHT) revolucionou o monitoramento contínuo da saúde. Entretanto, manipular dados de sensores de fontes distintas e relacioná-los com estados de saúde é uma tarefa complexa. Grafos de classificação são estruturas que podem auxiliar a manipulação desses dados, identificando relações entre sensores, algoritmos inteligentes e estados de saúde. No entanto, essas estruturas não são simples de interpretar e demandam visualizações adequadas, mas soluções desse tipo são pouco comuns. Este artigo apresenta a P-GRAFIT, uma plataforma projetada para auxiliar na interpretação, visualização e criação de grafos de classificação para sistemas IoHT. A solução atua como uma camada de abstração que converte dados complexos em representações visuais de forma automática, auxiliando desde a elicitação de requisitos até a seleção de algoritmos de Inteligência Artificial. Para avaliar a proposta, e comprovar a funcionalidade da plataforma, foi implementada uma prova de conceito com amostras de grafos de classificação pré-gerados. Além disso, exploramos um cenário de uso da plataforma para auxiliar o desenvolvimento de uma aplicação IoHT.

The increasing volume of unstructured texts in Official Gazettes highlights the need for robust semantic search engines. To address this, we propose a hybrid approach combining machine learning, human supervision, and a Large Language Model (LLM). The HALF (Human-Assisted Labeling Feedback) method, leveraging GPT-4o Mini, classified subjects in publications from the Official Gazette of Ceará. It assigned subjects to 1.044 publications with 0.8889 accuracy compared to ground truth. This approach enhances semantic search, improves retrieval and decision-making, and extends to other legal domains. Moreover, it offers a scalable solution, outperforming traditional unsupervised methods in accuracy and relevance.

Recommender Systems are information systems that provide personalized suggestions or recommendations to users based on their preferences, interaction history, and behavior patterns. An important use of recommender systems is in Virtual Learning Environments (VLEs), i.e. remote learning platforms where students can access educational materials, activities, assessments, etc., where they offer significant benefits to both students and instructors. Such benefits include personalized learning, increased engagement and effectiveness of both teaching and learning tasks.Challenges for the use of recommendation systems in VLEs include the cold start problem, data sparsity, and limited coverage.To deal with these challenges, we propose G-Learn, a recommendation system for contents in VLEs that can operate in both supervised and unsupervised modes. The recommendations are based on graph machine learning techniques combined with keyword mining and similarity, allowing the system to recommend educational materials that are adapted to the performance of each student detected when the student solves questions on the platform.We show the effectiveness of G-learn in a realistic scenario, using data from the VLE Homero, a computer science platform supported by the federal government of Brazil. We compare different settings for the keyword mining and machine learning techniques. Real data from students is used to validate the recommendation graphs, where G-learn obtained an average f1-score of 0.64 in the unsupervised mode and 0.95 in the supervised one.

In recent years, we have observed an increased volume of proposals for Internet of Health Things (IoHT) applications in the literature. Developing this type of application takes work, as it involves a series of challenges, many of them related to the constant change in the context of these applications, such as challenges related to energy consumption, the type of network communication, and its latency. In this sense, the development of self-adaptive systems is an alternative that has been explored for these applications. Self-adaptive systems can adapt to changes in context automatically, changing their behavior. However, developing self-adaptive systems is not trivial since, unlike other systems, they must be designed to evolve and adapt their behavior at runtime continuously. In this sense, we propose a Kotlin framework for developing self-adaptive IoHT applications for Android mobile devices called KREATION. This framework was developed using the Model-View-Control architecture and incorporates the MAPE-K adaptation loop into its internal logic. It also presents mechanisms to help collect data from Android smartphone sensors and data from the Google Fit API, which allows you to obtain data collected from other Android devices, such as smart bands and smartwatches. We evaluated the framework using it to develop two proofs of concept demonstrating that it is functional and can help the development of self-adaptive IoHT applications.

Information and communication technology networks are developing rapidly, and many users' daily activities, including e-banking and e-commerce, have moved to the internet. By using this anonymous framework, attackers establish new techniques such as phishing to trick victims into using fake sites to gather sensitive information. Detection of malicious sites is a challenging problem, due to its semantic-based attack framework, which exploits system and user vulnerabilities. While several approaches have been proposed to cope with phishing, new attacks constantly emerge requiring the improvement of existing techniques. This work proposes an approach to detect phishing, evaluating the structure of the graph from the connection of sites to extract features from the graph, in addition to the features commonly used in the literature such as features of URL and SSL certificate. We evaluate the impact of the use of graph features compared to existing features in the literature, from the point of view of machine learning metrics and of computational time required. The obtained results suggest that graph machine learning can be used effectively in phishing detection.

The Internet of Things (IoT) can be understood as a network of smart devices connected to the Internet that collect and share data. The Internet of Health Things (IoHT) is an area of research that has been gaining a lot of prominences and includes the use of IoT solutions aimed at monitoring healthcare and improving the health of users of these solutions. The development of IoHT solutions involves several challenges, including the interoperability between different smart devices and different sets of sensors, the difficulties of initial design decisions about which technology to use for the solution, and the development cost for the smart devices, which includes not only the financial cost but also the processing and energetic cost that are important to be considered, since the hardware limitations of these smart devices can impede the execution of some applications. The reuse of software artifacts can help reduce costs and other challenges. In this sense, this work proposes the modeling and implementation of a classification graph that relates different sensors, features, classification algorithms, and health states (or situations), providing a reusable software artifact that can help both the requirements elicitation and design stages. Also, the classification graph can be used as a knowledge base for implementing data analysis modules and predicting health states. We evaluated the proposal through a proof of concept, in which we implemented a classification graph based on the proposed model in a web server, using two different datasets, with data from accelerometer and gyroscope sensors and 30 actions (or final states). The developed system also implements three classification algorithms: an artificial neural network, a decision tree, and a random forest. Moreover, we developed a Web API to execute requests for both the creation and update of the classification graph and the request to download the optimized graph and the trained models by the classification algorithms based on the application requirements that requested the graph.

Internet access in both private and public environments allows users to broadly access their data what makes possible the deployment of new services based on Internet of Things. This fact created Smart Environments (SEs) that are composed of a huge amount of heterogeneous devices, for example, personal devices (smartphones, notebooks, tablets, etc) and IoT devices (sensors, actuators, and others). However, these environments can facilitate the action of malicious agents interested in promoting Distributed Denial of Service (DDoS) attacks to the network, and, when they are public places, it is challenging to locate these attackers. In this way, it is necessary to deploy solutions that can detect DDoS in SEs and to determine the physical location of the attacker, which is essential to prevent future attacks. Within this context, this article presents an Intelligent System for detection of DDoS and physical location of devices in SEs, applying Machine Learning (ML) and trilateration techniques. The experiments performed, using real network traffic and simulation, suggest that the proposed system is capable of detecting attacks and finding malicious devices.

This study proposes a development process for self-adaptive IoHT applications based on movement pattern data for mobile devices. The proposed process adapts the component reuse-oriented development process by adding elements related to the construction of self-adaptive systems. Also, we propose five artifacts to support this process: a taxonomy relating data from sensors and health status to assist; a correlation graph between sensor data, movement features, and health situations, to assist the choice of components and serving as a knowledge base for self-adaptive application; a template for building adaptation rules; a model based on the MAPE-K adaptation cycle; and a framework based on the model, for assisting the implementation phase. The evaluation of the proposed process will be done using proofs of concept and an experiment.

In the last two decades, studies about using technology for automatic detection of human falls increased considerably. The automatic detection of falls allows for quicker aid that is key to increasing the chances of treatment and mitigating the consequences of falls. However, each type of fall has its specificities and determining the correct type of fall can help treat the person who has fallen. Although it is essential to use computational methods to classify falls, there are few studies about that in the literature, especially compared to the studies that propose solutions for fall detection. In this sense, we execute a systematic literature review (SLR) using the (Kitchenham et al., 2009) method to investigate the computational solutions used to classify the different types of falls. We performed a search on Scopus, Web of Science, and PubMed scientific databases looking for computational methods to fall classification in their papers. We use the grounded theory methodology for a more detailed qualitative analysis of the papers. As a result of our search, we selected a total of 36 studies for our review and found two different computational methods for classifying falls. Related to the steps used in each method, we found fourteen different types of sensors, four different techniques for background and foreground extraction of videos, twenty-one techniques for feature extraction, and seven different fall classification strategies. Finally, we also identified fifty-one different types of falls. In conclusion, we believe that the methods and techniques analyzed in our study can help developers to create new and better systems for classification, detection, and prevention of falls and falls database. Besides, we identified gaps that can be explored in future research related to the automatic classification of falls.

The source code similarity problem consists in defining if two given distinct codes are the same program or not. This problem is valuable for polymorphic malware detection, which can generate distinct versions of code applying rules of obfuscation that change the original code. A technique to measure the similarity between source codes is to model the codes as program dependency graphs (PDG) and find an alignment between the graphs that maximizes a similarity function. This work investigates the performance of distinct similarity metrics applied to a genetic algorithm (GA). To distinguish between similar codes and non-similar codes, we use the k-nearest-neighbors (KNN) algorithm based on the similarity of the alignment found by the GA. The experiments are conducted with a database proposed by this work, where the source codes were retrieved from the Codeforces website and obfuscated by the tool CXX-obfuscator.

An effort is being made by authorities worldwide to improve the safety of the transportation of goods while preserving efficiency. Vehicle inspections are important for safety but not very frequent. When they do happen, vehicles are selected on the roadside and authorities spend a long time retrieving the relevant information while the vehicle is stopped. In this paper, we present and evaluate STOP, a road transportation vehicle inspection support system with tamper-proof records to prevent location spoofing attacks. To the best of our knowledge, it is the first such system described in literature. The STOP system uses mobile devices and a central server to allow authorities to select and notify vehicles for inspection while retrieving the needed information to prepare the procedure beforehand. The location chain for each vehicle can be verified and signed by the inspectors. We implemented a prototype in the Android platform and tested it with real users. We evaluated the system’s location retrieval accuracy, response times, and Bluetooth communication during inspection.

We investigate graph colouring models for the purpose of optimizing TDMA link scheduling in Wireless Networks. Inspired by the BPRN-colouring model recently introduced by Rocha and Sasaki, we introduce a new colouring model, namely the BMRN-colouring model, which can be used to model link scheduling problems where particular types of collisions must be avoided during the node transmissions. In this paper, we initiate the study of the BMRN-colouring model by providing several bounds on the minimum number of colours needed to BMRN-colour digraphs, as well as several complexity results establishing the hardness of finding optimal colourings. We also give a special focus on these considerations for planar digraph topologies, for which we provide refined results.

A Dynamic Software Product Line (DSPL) allows the generation of products that can adapt dynamically according to changes in requirements or environment at runtime. This runtime adaptation is often made by the activation and deactivation of features, introducing a cost (e.g., an overhead regarding resource consumption). To reduce this cost, a solution is the partial product configuration at the static binding time. Thus, in DSPLs, one challenge is the feature selection to define which features should be bound permanently before the initial launch and which features should be bound at runtime. In this paper, we address this challenge presenting a graph model formulation to the feature selection problem for the initial launch in DSPLs that considers both static and dynamic binding. This model allows the application of efficient optimization algorithms to solve the problem. We also present a proof of concept showing that the model can be used to generate optimized solutions to the feature selection problem for initial launch in DSPLs.

DEA (Data Envelopment Analysis) is a performance evaluation technique that allows evaluating relative efficiency between similar productive units.Efficient units form the efficiency frontier.Inefficient units use efficient units as a reference to reach the efficiency frontier, making DEA a benchmarking technique.In a practical application of the technique, the transition from a very inefficient unit to the efficiency frontier may not be feasible in one step.It is necessary to create mechanisms of intermediate goals for efficiency to be achieved.In this paper we propose an approach that involves the creation of a directed weighted graph whose vertices are the benchmarks of each step for transforming inefficient units into efficient ones.Through a minimum path in this graph, it is possible to indicate a smaller sequence of steps to make an inefficient unit to become efficient.

A radio network consists of a set of transceiver nodes in space that communicate using broadcast radio. Since communication is done over a shared medium, transmissions are subject to collisions. Different Medium Access Control techniques are used to avoid such collisions and subsequent data loss. In this article, we study Time Division Multiple Access link scheduling in Wireless Multihop Networks. We generalize the packet radio network (PRN)‐coloring model that was used in previous works to obtain the Backbone PRN (BPRN)‐coloring. The BPRN‐coloring captures the fact that typically only a subset of links need to be scheduled, corresponding to the backbone network. We study the BPRN‐coloring and the corresponding BPRN‐chromatic index considering a rooted tree as backbone, motivated by applications in Wireless Sensor Networks. The BPRN‐chromatic index is determined when the whole graph is either a complete graph or a cycle, and we give partial results in the case of a bipartite graph. We show that determining the BPRN‐chromatic index is NP‐hard even when the network graph is bipartite, and the backbone is an oriented tree toward a root vertex. Finally, we model a ring topology as the power of a cycle graph and give an upper bound on the BPRN‐chromatic index. © 2017 Wiley Periodicals, Inc. NETWORKS, Vol. 71(4), 403–411 2018

A radio network consists of a set of transceiver nodes that make use of radio transmissions to communicate with each other. Because these wireless networks share the communication channel, collisions may occur in transmissions, whether through primary or secondary interference. Considering that to a large extent these networks have multi-hops in their composition, thus it is possible the spatial reuse in the channel sharing, with the prevention of collisions and with this reducing the loss of data transmitted. In the context of these networks, several approaches are used to medium control access, in order to coordinate access to the wireless channel to avoid overlapping of transmissions in the use of the channel. A wireless network can be represented as a digraph, and the PRN (Packet Radio Network) coloring model is applied as a scheduling criterion for MAC protocols. The BPRN (Backbone PRN) model represents a generalization of the PRN coloring, applying a more realistic approach where it is considered only a subset of links that will be colored, which we call the backbone network. The backbone network considers the fact that in a multi-hop wireless network, a transceiver node communicates only with some of its neighbors. In this work, the backbone network is represented by a tree oriented towards a root node, due to its relationship to the domain of the wireless sensor networks. Thus, this work proposes the development of algorithms for coloring graphs based on the BPRN coloring model, with emphasis on the TDMA channel allocation technique, since it will be used for the scheduling of links in the context of multi-hop wireless networks.

Neste artigo é apresentado o protocolo SR-MAC, um protocolo crosslayer baseado em coloração de grafos para redes de sensores sem fio. A principal característica do protocolo é um uso eficiente dos intervalos de transmissão alocados aos sensores, por meio de um escalonamento TDMA, oferecendo escalabilidade e eficiência na transmissão dos dados coletados para a estação base. O SR-MAC foi validado por meio de um simulador específico para redes de sensores. O protocolo apresentou bons resultados quando comparado a protocolos que não fazem reuso dos intervalos de transmissão, mesmo em cenários com alta densidade, sendo capaz de suportar 53% mais nós e possuindo uma eficiência energética na coleta de dados superior em mais de 30%.

Wireless Sensor Networks (WSN) consist in a set ofsensor nodes that collect data in the environment and send it to a Base Station that processes the final data. Some challenges may be found, such as minimizing energy consumption and maximize the network lifetime. Many protocols achieve energy savings through network clustering. This paper presents a new modeling graph using flow network to improve the routing protocols using the clustering technique, improving the routing of data between clusters-head, as well as by setting a time for a round in the clustering process. With this agnostic technique, our proposal may increase the network lifetime, balance the power consumption among the nodes and better distribute the data transmissions. Through simulations in network simulator Omnet++, Castalia, the proposal is validated and its efficiency is verified by comparing it to other protocols which work with clustering techniques.

This paper describes an approach to the initialization of Multi-Objective Genetic Algorithms (MOGA). The proposed approach inserts in the initial population some solutions that are already in the Pareto optimal front or near it. These are extreme solutions, and a set of conveniently spaced solutions in the Pareto optimal front, obtained by exact algorithms or heuristics over a mono-objective formulation of the problem. To complete the initial population, the algorithm constructs a path connecting these solutions using an algorithm based on PathRelinking. The performance of this boot approach is compared against the random initialization, the insertion of optimal or sub-optimal solutions without the use of the PathRelinking, and some initialization heuristics that are problem-specific. The results of the empirical comparison provide clear evidence that supports the conclusion that the proposed approach is better than the others in terms of overall effectiveness.