REPOSITÓRIO P.PORTO
Scientific Repository of the Porto Polytechnic
Recent Submissions
IoT intrusion detection through machine learning: Research Group on Intelligent Engineering and Computing for Advanced Innovation and Development: 2020-2021
Publication . Vitorino, João; Sousa, Orlando; Praça, Isabel; Andrade, Rui
The digital transformation faces great security challenges. In particular, the Internet of Things (IoT), a concept that expresses the interconnection of physical objects with the Internet, is exposed to several threats. The growing number of cyber attacks targeting IoT systems restates the need for a reliable detection of malicious network activity, to mitigate its impact. The application of Machine Learning (ML) to IoT intrusion detection is a promising approach to tackle the increasingly more complex threats.
This project presents a continuously improving Network-based Intrusion Detection System (NIDS) based on user feedback. The system consists of three modular applications and employs an adapted Deep Reinforcement Learning (DRL) methodology to incrementally improve the detection with the alerts validated by end users.
The binary and multi-class classification performances of the developed DRL model, a Support Vector Machine (SVM), a Light Gradient Boosting Machine (LightGBM), an Extreme Gradient Boosting (XGBoost), an Isolation Forest (iForest) and a Local Outlier Factor (LOF) were evaluated on several subsets of the IoT-23 dataset.
The obtained results indicate that the DRL model requires a large quantity of balanced data to be effective, whereas iForest and LOF are more suitable for a smaller quantity of unbalanced data. Overall, SVM, LightGBM and XGBoost obtained similar results. LightGBM achieved the most reliable performance.
Quebrando estruturas de vidro: A mulher nas organizações
Publication . Menezes, Mariana; Barbosa Gomes Pereira, Pedro Duarte
Apesar da luta pela igualdade de género a que se tem assistido ao longo das últimas décadas, e de todas as conquistas da mulher, as posições de liderança e gestão nas organizações continuam a ser maioritariamente ocupadas por homens. Aquando da sua ascendência na hierarquia organizacional, a mulher atinge um “telhado de vidro” que condiciona o seu sucesso profissional e coloca em causa o equilíbrio entre a sua vida pessoal e profissional. Isto porque as organizações não são neutras em termos de género, possuindo estereótipos desta índole integrados estruturalmente, perpetuados através da comunicação explícita e implícita das organizações. Para combater a segregação vertical e horizontal, é fundamental a legislação e medidas organizacionais, bem como a educação e formação dos indivíduos, de forma a promover a desconstrução de estereótipos e preconceitos associados à mulher no mundo organizacional. Assim, o presente trabalho argumentará como a desigualdade de género está presente no seio das organizações.
Vitamin K Epoxide Reductase Complex (VKORC1) Electrochemical Genosensors: Towards the Identification of 1639 G>A Genetic Polymorphism
Publication . Barbosa, Tiago; Morais, Stephanie L.; Carvalho, Renato; Magalhães, Júlia M. C. S.; Domingues, Valentina F.; Delerue-Matos, Cristina; Ferreira-Fernandes, Hygor; Pinto, Giovanny R.; Santos, Marlene; Barroso, Maria Fátima; Santos, Marlene; Manuel de Caldevilla Carvalho, Renato
Anticoagulants, including warfarin, are often administered to patients who are exhibiting early symptoms of thromboembolic episodes or who have already experienced such episodes. However, warfarin has a limited therapeutic index and might cause bleeding and other clinical problems. Warfarin inhibits the vitamin K epoxide reductase complex subunit 1 (VKORC1), an enzyme essential for activating vitamin K, in the coagulation cascade. Genetic factors, such as polymorphisms, can change the natural function of VKORC1, causing variations in the medication reaction among individuals. Hence, before prescribing warfarin, the patient’s genetic profile should also be considered. In this study, an electrochemical genosensor capable of detecting the VKORC1 1639 G>A polymorphism was designed and optimized. This analytical approach detects the electric current obtained during the hybridization reaction between two 52 base pair complementary oligonucleotide sequences. Investigating public bioinformatic platforms, two DNA sequences with the A and G single-nucleotide variants were selected and designed. The experimental protocol of the genosensor implied the formation of a bilayer composed of a thiolate DNA and an alkanethiol immobilized onto gold electrodes, as well as the formation of a DNA duplex using a sandwich-format hybridization reaction through a fluorescein labelled DNA signalling probe and the enzymatic amplification of the electrochemical signal, detected by chronoamperometry. A detection limit of 20 pM and a linear range of 0.05–1.00 nM was obtained. A clear differentiation between A/A, G/A and G/G genotypes in biological samples was successfully identified by his novel device.
Data Quality Assessment: A Practical Application
Publication . Eliana Costa e Silva; Teresa Peixoto; Óscar Oliveira; Bruno Oliveira
This paper presents a novel data quality score designed to address the challenges of ensuring high-quality data in Internet of Things (IoT) deployments. Given the growing reliance on IoT systems and the
volume of data they generate, maintaining data quality is essential for reliable decision-making and effective analytics. The proposed score synthesizes key data quality dimensions, providing a comprehensive measure of data quality that can be applied across various IoT contexts. The results obtained for a public dataset on a water pumping system show the applicability and flexibility of the proposed data quality score. This work contributes to the ongoing efforts to improve data management in IoT environments, ultimately supporting the development of robust, data-driven solutions.
Understanding musculoskeletal loadings among supermarket checkout counter cashiers: A biomechanical analysis
Publication . Silva, Tânia T.; Sousa, Catarina; Colim, Ana; Rodrigues, Matilde A.
Work-related musculoskeletal disorders (WMSD) are highly prevalent among supermarket cashiers. These disorders are frequently related to the adoption of awkward postures and manual materials handling. This study aimed to analyze musculoskeletal loadings in supermarket cashiers, considering the handling of different products and different checkout conditions. To accomplish this, we employed an inertial motion capture system to measure full-body kinematics while simulating 19 cashier tasks. The study included five female cashiers from a supermarket in Northern Portugal, ranging in age from 19 to 61 years old. Using joint angles, material load, and muscle function as input parameters, we conducted the musculoskeletal loadings assessment using the Rapid Upper Limb Assessment (RULA) and Rapid Entire Body Assessment (REBA) methods. Results showed that RULA scores were higher for the microtasks that involved product scanning. Regarding microstasks analyzed by REBA, the replacement of paper rolls for the receipt machine at the checkout counter yields the highest scores. Based on these findings, there is a compelling need to redesign supermarket checkout workstations to alleviate the physical demands placed on cashiers and to ensure organizational sustainability.