Pasar al contenido principal

GESTION INTELIGENTE DEL CLOUD CONTINUUM: DESARROLLO DE LAS FUNCIONALIDADES CLAVE DE UN SO

PID2021-124463OB-I00

•
Nombre agencia financiadora Agencia Estatal de Investigación
Acrónimo agencia financiadora AEI
Programa Programa Estatal para Impulsar la Investigación Científico-Técnica y su Transferencia
Subprograma Subprograma Estatal de Generación de Conocimiento
Convocatoria Proyectos de I+D+i (Generación de Conocimiento y Retos Investigación)
Año convocatoria 2021
Plan Estatal 2021-2023 Plan Estatal de Investigación Científica y Técnica y de Innovación
Centro beneficiario UNIVERSITAT POLITECNICA DE CATALUNYA

Resultados relacionados

Resultados totales (Incluyendo duplicados): 41
Encontrada(s) 1 página(s)

UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Analyzing distinct neural network models for oxygen saturation prediction towards a personalized COPD management

  • Pascual Saldaña, Heribert|||0000-0003-2114-0929
  • Masip Bruin, Xavier|||0000-0002-4755-556X
  • Alonso Beltran, Albert|||0000-0003-0921-8033
  • Blanco Vich, Isabel
Chronic Respiratory Diseases (CRDs), including Chronic Obstructive Pulmonary Disease (COPD), pose significant global health challenges. Long-term oxygen therapy (LTOT) has been widely adopted to improve respiratory function and quality of life in CRD patients. However, current oxygen dosing approaches show several limitations, including inconvenience and mismatched oxygen flow rates according to patients' activity. To address these challenges, the design of Patient Specific Modeling (PSM) and the utilization of advanced technologies, such as artificial intelligence (AI), can be integrated towards a personalized oxygen dosing. This study presents a first draft of a conceptual architecture combining cloud and edge computing aimed at optimizing oxygen dosing based on individual patient data. The evaluation of sixty neural network architectures in a small-scale test has shown good results, with smaller networks generally outperforming larger ones. Further research is needed to design new testing procedures and explore optimal parameters for the neural networks, including a large patients' number., This work has been supported by the Spanish Ministry of Science and Innovation under grant PID2021-124463OB-I00 funded by MCIN/AEI/10.13039/501100011033 and by ERDF A way of making Europe, the Catalan Government under contract 2021 SGR 00326 and the Catalan Department of Research and Universities, Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

A 3D terrain generator: Enhancing robotics simulations with GANs

  • Arellano García, Silvia
  • Otero Calviño, Beatriz|||0000-0002-9194-559X
  • Kucner, Tomasz Piotr
  • Canal Corretger, Ramon|||0000-0003-4542-204X
Simulation is essential in robotics to evaluate models and techniques in a controlled setting before conducting experiments on tangible agents. However, developing simulation environments can be a challenging and time-consuming task. To address this issue, a proposed solution involves building a functional pipeline that generates 3D realistic terrains using Generative Adversarial Networks (GANs). By using GANs to create terrain, the pipeline can quickly and efficiently generate detailed surfaces, saving researchers time and effort in developing simulation environments for their experiments. The proposed model utilizes a Deep Convolutional Generative Adversarial Network (DCGAN) to generate heightmaps, which are trained on a custom database consisting of real heightmaps. Furthermore, an Enhanced Super-Resolution Generative Adversarial Network (ESRGAN) is used to improve the resolution of the resulting heightmaps, enhancing their visual quality and realism. To generate a texture according to the topography of the heightmap, chroma keying is used with previously selected textures. The heightmap and texture are then rendered and integrated, resulting in a realistic 3D terrain. Together, these techniques enable the model to generate high-quality, realistic 3D terrains for use in robotic simulators, allowing for more accurate and effective evaluations of robotics models and techniques., This work is partially supported by the Spanish Ministry of Science and Innovation under contract PID2021-124463OB-IOO, by the Generalitat de Catalunya under grant 2021-SGR-00326. Finally, the research leading to these results also has received funding from the European Union’s Horizon 2020 research and innovation programme under the HORIZON-EU VITAMIN-V (101093062) project., Peer Reviewed, Objectius de Desenvolupament Sostenible::9 - Indústria, Innovació i Infraestructura
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Service placement in the continuum: A systematic literature review

  • Sajjad, Waseem|||0009-0007-4956-6995
  • Farreras Esclusa, Montserrat|||0000-0003-2027-1919
  • García Almiñana, Jordi|||0000-0002-3515-7150
  • Masip Bruin, Xavier|||0000-0002-4755-556X
Cloud computing plays a crucial role in the Industry 4.0 era, particularly with the rise of Internet of Things (IoT) applications that support domains such as education, healthcare, business, and manufacturing. These applications consist of multiple services with diverse quality of service (QoS) requirements, making their development and deployment complex. While traditional cloud environments provide scalability, they often fail to support latency-sensitive and resource-intensive applications. To overcome these limitations, alternative paradigms such as Cloud–Fog–Edge (CFE), Cloud–Fog (CF), Cloud–Edge (CE), Fog–Edge (FE), and Mobile Edge Computing (MEC) have emerged. These models push computation, storage, and networking closer to end devices, reducing latency and bandwidth usage. However, the heterogeneity, mobility, and dynamic nature of these environments make service placement (a known NP-hard problem) a central challenge. This article presents a systematic literature review of service placement approaches across the compute continuum. Following established SLR methodology, we identified and analyzed 124 peer-reviewed studies published between 2018 and 2024, classifying them by (i) deployment environment, (ii) service placement strategies and algorithms, (iii) adaptability of the solution, (iv) optimization objectives, (v) virtualization/orchestration technologies, (vi) evaluation methodologies, including workloads, testbeds, and simulation tools and (vii) use cases or application types. The novelty of this work lies in providing not only a detailed taxonomy of placement approaches but also this is the first survey that takes all seven aspects into consideration and establishes correlations between them. Our findings reveal that most existing works target smart health applications and favor heuristic-based placement in complex CFE scenarios, while research on scientific and compute-intensive workloads remains limited. We also identify Kubernetes as the most widely used orchestration technology and latency as the dominant optimization metric. Despite significant progress, the field is still maturing, with gaps in real-world validation and adaptive, ML-based placement strategies. By consolidating technical approaches, evaluation practices, and open challenges, this survey offers both researchers and practitioners a structured overview of the state of the art and guidance for advancing service placement in the compute continuum., The work leading to this publication was partially co-funded by the Spanish Ministry of Science and Innovation under grant PID2021-124463OB-I00 funded by MCIN /AEI /10.13039/501100011033 and by the ERDF ‘‘A way of making Europe’’; by the Catalan Government under contract 2021 SGR 00326; and by the EU Horizon programme under grant 101070177 (ICOS project)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

A machine-learning-based framework for detection and recommendation in response to cyberattacks in critical energy infrastructures

  • Rabadan Arroyo, Raul
  • Hussain, Ayaz|||0000-0002-9081-9567
  • Simó Mezquita, Ester|||0000-0003-4587-7028
  • Rodríguez Luna, Eva|||0000-0001-5904-7039
  • Masip Bruin, Xavier|||0000-0002-4755-556X
This paper presents an attack detection, response, and recommendation framework designed to protect the integrity and operational continuity of IoT-based critical infrastructure, specifically focusing on an energy use case. With the growing deployment of IoT-enabled smart meters in energy systems, ensuring data integrity is essential. The proposed framework monitors smart meter data in real time, identifying deviations that may indicate data tampering or device malfunctions. The system comprises two main components: an attack detection and prediction module based on machine learning (ML) models and a response and adaptation module that recommends countermeasures. The detection module employs a forecasting model using a long short-term memory (LSTM) architecture, followed by a dense layer to predict future readings. It also integrates a statistical thresholding technique based on Tukey’s fences to detect abnormal deviations. The system was evaluated on real smart meter data in a testbed environment. It achieved accurate forecasting (MAPE < 2% in most cases) and successfully flagged injected anomalies with a low false positive rate, an effective result given the lightweight, unsupervised, and real-time nature of the approach. These findings confirm the framework’s applicability in resource-constrained energy systems requiring real-time cyberattack detection and mitigation., This work has been partially supported by the HE PHOENI2X project, funded by European Commission with Grant Number 101070586, and by the Spanish Ministry of Science and Innovation under grant PID2021-124463OB-I00, funded by MCIN/AEI/10.13039/501100011033 and by ERDF A way of making Europe, as well as by the Catalan Government under contract 2021 641 SGR 00326., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

A deep learning framework for safety monitoring of a railway section

  • Chriki Zerrouk, Fatima Zohra
  • Simó Mezquita, Ester|||0000-0003-4587-7028
  • Aguiló Gost, Francisco de Asis L.|||0000-0003-1572-8486
  • Garcia Milà, Ignasi
  • Masip Bruin, Xavier|||0000-0002-4755-556X
© 2024 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works., This paper presents an approach for anomaly detection and forecasting that aims to protect the integrity and operational continuity of IoT critical infrastructure, specifically for a railway use case. This system is designed to monitor in real-time the sensor data to detect any deviations that may indicate a potential data tampering attack or a sensor malfunction. This early identification and notification of such anomalies is crucial for preventing unauthorized access and mitigating the risks associated with data tampering attacks. The proposed system is comprises of two primary components: a forecasting component and an anomaly detection component. The Forecasting Component uses a Long Short-Term Memory (LSTM) model to predict future sensor values, while the Anomaly Detection Component employs the Tukey's fence method to identify sensor measurements that significantly deviate from normal behaviour. A railway use case is included to demonstrate the practical application of the deep learning framework. These components were evaluated and both demonstrated excellent performance. The Forecasting Component provided highly accurate predictions of future sensor values, while the Anomaly Detection Component effectively identified deviations from normal patterns. The evaluation results confirmed the system's ability to detect significant anomalies and its capability to maintain operational integrity and security in IoT critical infrastructures., The work leading to this publication was partly funded by the European Union’s Horizon Europe under project PHOENi2X, Grant Agreement no. 101070586. For UPC authors this work is partially co-funded by the Spanish Ministry of Science and Innovation under grant PID2021-124463OBI00 funded by MCIN/AEI/10.13039/501100011033 and by ERDF A way of making Europe, as well as by the Catalan Government under contract 2021 SGR 00326., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

A comprehensive architecture for federated learning-based smart advertising

  • Seyghaly, Rasool|||0000-0002-2805-5607
  • García Almiñana, Jordi|||0000-0002-3515-7150
  • Masip Bruin, Xavier|||0000-0002-4755-556X
This paper introduces a cutting-edge data architecture designed for a smart advertising context, prioritizing efficient data flow and performance, robust security, while guaranteeing data privacy and integrity. At the core of this study lies the application of federated learning (FL) as the primary methodology, which emphasizes the authenticity and privacy of data while promptly discarding irrelevant or fraudulent information. Our innovative data model employs a semi-random role assignment strategy based on a variety of criteria to efficiently collect and amalgamate data. The architecture is composed of model nodes, data nodes, and validator nodes, where the role of each node is determined by factors such as computational capability, interconnection quality, and historical performance records. A key feature of our proposed system is the selective engagement of a subset of nodes for modeling and validation, optimizing resource use and minimizing data loss. The AROUND social network platform serves as a real-world case study, illustrating the efficacy of our data architecture in a practical setting. Both simulated and real implementations of our architecture showcase its potential to dramatically curtail network traffic and average CPU usage, while preserving the accuracy of the FL model. Remarkably, the system is capable of achieving over a 50% reduction in both network traffic and average CPU usage even when the user count escalates by twenty-fold. The click rate, user engagement, and other parameters have also been evaluated, proving that the proposed architecture’s advantages do not affect the smart advertising accuracy. These findings highlight the proposed architecture’s capacity to scale efficiently and maintain high performance in smart advertising environments, making it a valuable contribution to the evolving landscape of digital marketing and FL., This work has been supported by the Spanish Ministry of Science and Innovation under grant PID2021-124463OB-I00 funded by MCIN/AEI/10.13039/501100011033 and by ERDF A way of making Europe, and by the Catalan Government under contract 2021 SGR 00326. The corresponding author, R.S., gratefully acknowledges the Universitat Politècnica de Catalunya and Banco Santander for the financial support of his predoctoral grant FPI-UPC 2021, Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Low-cost training of image-to-image diffusion models with incremental learning and task/domain adaptation

  • Antona Pizà, Héctor
  • Otero Calviño, Beatriz|||0000-0002-9194-559X
  • Tous Liesa, Rubén|||0000-0002-1409-5843
Diffusion models specialized in image-to-image translation tasks, like inpainting and colorization, have outperformed the state of the art, yet their computational requirements are exceptionally demanding. This study analyzes different strategies to train image-to-image diffusion models in a low-resource setting. The studied strategies include incremental learning and task/domain transfer learning. First, a base model for human face inpainting is trained from scratch with an incremental learning strategy. The resulting model achieves an FID score almost equivalent to that of its batch learning equivalent while significantly reducing the training time. Second, the base model is fine-tuned to perform a different task, image colorization, and, in a different domain, landscape images. The resulting colorization models showcase exceptional performances with a minimal number of training epochs. We examine the impact of different configurations and provide insights into the ability of image-to-image diffusion models for transfer learning across tasks and domains., This work is partially supported by the Spanish Ministry of Science and Innovation under contracts PID2019-107255GB and PID2021-124463OB-IOO, by the Generalitat de Catalunya under grants 2017-SGR-962, 2021-SGR-00326, 2021-SGR-00478 and DRAC (IU16-011591). The research leading to these results has received funding from the European Union’s Horizon 2020 research and innovation program under the HORIZON-AG PHOENI2X (101070586) and HORIZON-EU VITAMIN-V (101093062) projects., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Automatic Linux malware detection using binary inspection and runtime opcode tracing

  • Alonso García, Martí
  • Gironès, Andreu
  • Costa Prats, Juan José|||0000-0003-2479-0230
  • Morancho Llena, Enrique|||0000-0003-2403-8145
  • Di Carlo, Stefano
  • Canal Corretger, Ramon|||0000-0003-4542-204X
The fast-paced evolution of cyberattacks to digital infrastructures requires new protection mechanisms to counterattack them. Malware attacks, a type of cyberattacks ranging from viruses and worms to ransomware and spyware, have been traditionally detected using signature-based methods. But with new versions of malware, this approach is not good enough, and new machine learning tools look promising. In this paper we present two methods to detect Linux malware using machine learning models: (1) a dynamic approach, that tracks the application executed instructions (opcodes) while they are being executed; and (2) a static approach, that inspects the binary application files before execution. We evaluate (1) five machine learning models (Support Vector Machine, k-Nearest Neighbor, Naive Bayes, Decision Tree and Random Forest) and (2) a deep neural network using a Long Short-Term Memory architecture with word embedding. We show the methodology, the initial dataset preparation, the infrastructure used to obtain the traces of executed instructions, and the evaluation of the results for the different models used. The obtained results show that the dynamic approach with a Random Forest classifier gets a 90% accuracy or higher, while the static approach obtains a 98% accuracy., We thank the anonymous reviewers for their thoughtful comments and suggestions which have improved the presentation of this work. This work is partially supported by the Spanish Ministry of Science and Innovation under contract PID2021-124463OB-IOO and by the Generalitat de Catalunya, Spain under grant 2021-SGR-00326. Finally, the research leading to these results has received funding from the European Union’s Horizon Programme under the HORIZON-EU VITAMIN-V (101093062) project., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

SBNNR: small-size bat-optimized KNN regression

  • Seyghaly, Rasool|||0000-0002-2805-5607
  • García Almiñana, Jordi|||0000-0002-3515-7150
  • Masip Bruin, Xavier|||0000-0002-4755-556X
  • Kuljanin, Jovana|||0000-0002-3380-262X
Small datasets are frequent in some scientific fields. Such datasets are usually created due to the difficulty or cost of producing laboratory and experimental data. On the other hand, researchers are interested in using machine learning methods to analyze this scale of data. For this reason, in some cases, low-performance, overfitting models are developed for small-scale data. As a result, it appears necessary to develop methods for dealing with this type of data. In this research, we provide a new and innovative framework for regression problems with a small sample size. The base of our proposed method is the K-nearest neighbors (KNN) algorithm. For feature selection, instance selection, and hyperparameter tuning, we use the bat optimization algorithm (BA). Generative Adversarial Networks (GANs) are employed to generate synthetic data, effectively addressing the challenges associated with data sparsity. Concurrently, Deep Neural Networks (DNNs), as a deep learning approach, are utilized for feature extraction from both synthetic and real datasets. This hybrid framework integrates KNN, DNN, and GAN as foundational components and is optimized in multiple aspects (features, instances, and hyperparameters) using BA. The outcomes exhibit an enhancement of up to 5% in the coefficient of determination (¿^2 score) using the proposed method compared to the standard KNN method optimized through grid search., This work has been supported by the Spanish Ministry of Science and Innovation under grant PID2021-124463OB-I00, funded by MCIN/AEI/10.13039/501100011033 and by ERDF “A way of making Europe”, by the European Union’s Horizon Europe under the HE ICOS project, Grant Agreement no. 101070177, and by the Catalan Government under contract 2021 SGR 00326. The corresponding author R.S. gratefully acknowledges the Universitat Politècnica de Catalunya and Banco Santander for the financial support of his predoctoral grant FPI-UPC 2021., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

A security services management architecture toward resilient 6G wireless and computing ecosystems

  • Rodríguez Luna, Eva|||0000-0001-5904-7039
  • Masip Bruin, Xavier|||0000-0002-4755-556X
  • Martrat, Josep
  • Díaz, Rodrigo
  • Jukan, Admela
  • Granelli, Fabrizio
  • Trakadas, Panagiotis
  • Xilouris, George
The upcoming sixth-generation (6G) networks are expected to be more heterogeneous, scalable, reliable, secure and energy-efficient. 6G services and applications will benefit from the fast data transmission rates, global coverage, precise positioning, and widespread intelligence capabilities provided by 6G technology. In this complex scenario, new attack surfaces and vectors will emerge, due to the distributed, disaggregated, dynamic, programmable and open nature of the entire end-to-end 6G infrastructure along with the fragmentation of data, as well as the need for supporting cross-platforms interoperability, making the application of security and trust fundamentally challenging. To securely manage services in the future 6G mobile networks along with the set of functions ruling its management, conceptual work is needed to defining functional blocks supporting a secure end-to-end 6G systems management. This is especially critical in handling specific functionalities, such as network disaggregation, risk and threats business impact, energyefficiency, self-configuration or dynamic discovery. This paper positions an innovative 6G security services management architecture, which builds on a set of innovative building blocks, leveraging key technologies, such as AI-assisted models and Digital Twin, while enabling a human-centric approach toward an end-to-end security solution. Two canonical workflows along with two illustrative application scenarios are proposed, Light Rail Transit and Extended Reality, to conceptually highlight the proposed architecture impact and its expected benefits when high bandwidth, ultra-low latency, and secure communications are required., This work was supported by the European union’s Horizon Europe, under the project HORSE, Grant 101096342. The work of Eva Rodriguez and Xavi Masip-Bruin was supported in part by the Spanish Ministry of Science and Innovation funded by MCIN/AEI/10.13039/501100011033 and ERDF: A way of making Europe under Grant PID2021-124463OB-I00, and in part by the Catalan Government under Contract 2021 SGR 00326., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

A survey on 5G private and B5G network threats and safeguarding AI-based security mechanisms through the layered analysis

  • Tariq, Saman|||0000-0001-7516-4008
  • Rodríguez Luna, Eva|||0000-0001-5904-7039
  • Masip Bruin, Xavier|||0000-0002-4755-556X
  • Díaz Rodríguez, Rodrigo
  • Martrat, Josep
  • Trakadas, Panagiotis
The fifth-generation (5G) mobile network has shifted the paradigm in connectivity, high-speed data transmission, ultra-low latency, ultra-high throughput, multisense transmission, and ultra-high reliability. Today, industries are increasingly adopting 5G private networks, which also handle sensitive data related to business information, trade secrets, and personal data. Attacks on 5G private networks can potentially result in losing a competitive advantage since different security threats and vulnerabilities progressively target these networks. The unique infrastructure of the 5G network architecture and key enabling technologies exposes them to various vulnerabilities that attackers can target to breach sensitive data, steal information, and disrupt critical systems. Therefore, paying special attention to the security issues of 5G private networks is essential. The advancement of future wireless technology came about because of the different and diverse nature of connected devices, in contrast to the previous generation of mobile networks. The B5G network provides support to open network platforms, open interfaces, and the integration of different key enabling technologies helps manage network services and deploy new services needed for diverse requirements. At the same time, it has increased the attack surface compared to previous-generation networks. Therefore, it is imperative to conduct a review that focuses on addressing and classifying the different emerging threats in the private 5G and B5G networks in a distinctive way. In this paper, we have adopted a layered architecture from an industrial use case of 5G private networks to identify and classify different threats in 5G private networks. The study also characterized and modeled the different threats using information on the type of attack, entry points, and impact of the attack on the architecture layer. Moreover, an analysis of key enablers in 5G private and B5G networks and information on security threats and cyber-attacks is also presented. To accommodate the emerging threats in next-generation wireless technology, we have classified and modeled the different threats in the B5G domain using the common involved layer, which includes perception, network, and application layers in Hexa-x E2E and 6G IoT-enabled architecture. Organizations and projects that do not actively engage in cutting-edge technologies will lose their competitive advantage in the evolving technological landscape. This study has mapped the identified threat categories with different types of threats that could occur at different layers and assessed the entry points of the attacker. We have also identified the attack’s impact at each layer using security requirements related to confidentiality, availability, and integrity. Furthermore, this study reviews different AI-enabled solutions that can add value in preserving the security of 5G and B5G networks. Through this review analysis, we can conclude that the layered approach is quite beneficial in identifying the different threats and what security solutions can be offered to mitigate them., The work leading to this publication was partly funded by the European Union’s Horizon Europe under the project HORSE, Grant Agreement no. 101096342. For UPC authors this work is partially cofunded by the Spanish Ministry of Science and Innovation under grant PID2021-124463OB-I00 funded by MCIN/AEI/10.13039/501100011033 and by ERDF A way of making Europe, as well as by the Catalan Government under contract 2021 SGR 00326., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Innovative predictive approach towards a personalized oxygen dosing system

  • Pascual Saldaña, Heribert|||0000-0003-2114-0929
  • Masip Bruin, Xavier|||0000-0002-4755-556X
  • Asensio Garcia, Adrian|||0000-0002-3345-6093
  • Alonso Beltran, Albert|||0000-0003-0921-8033
  • Blanco Vich, Isabel
Despite the large impact chronic obstructive pulmonary disease (COPD) that has on the population, the implementation of new technologies for diagnosis and treatment remains limited. Current practices in ambulatory oxygen therapy used in COPD rely on fixed doses overlooking the diverse activities which patients engage in. To address this challenge, we propose a software architecture aimed at delivering patient-personalized edge-based artificial intelligence (AI)-assisted models that are built upon data collected from patients’ previous experiences along with an evaluation function. The main objectives reside in proactively administering precise oxygen dosages in real time to the patient (the edge), leveraging individual patient data, previous experiences, and actual activity levels, thereby representing a substantial advancement over conventional oxygen dosing. Through a pilot test using vital sign data from a cohort of five patients, the limitations of a one-size-fits-all approach are demonstrated, thus highlighting the need for personalized treatment strategies. This study underscores the importance of adopting advanced technological approaches for ambulatory oxygen therapy., This research was funded by the Spanish Ministry of Science, Innovation and Universities and FEDER, grant number PID2021-124463OB-100, and by the AGAUR Catalan Agency, grant number 2021_SGR_00326., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Enhancing cybersecurity in railways: Machine learning approaches for attack detection

  • Otero Calviño, Beatriz|||0000-0002-9194-559X
  • Rodríguez Luna, Eva|||0000-0001-5904-7039
  • Costa Prats, Juan José|||0000-0003-2479-0230
  • Oriol, Mercedes
Ensuring the security of railway systems is crucial to protecting both passengers and infrastructure from the increasing threat of cyberattacks. As these threats grow in complexity and frequency, the need for resilient attack detection systems becomes more pressing. This study tackles the challenge of attack detection in railway systems using machine learning techniques. By integrating Generative Adversarial Networks (GANs), Convolutional Neural Networks (CNNs), and Transfer Learning (TL), we enhance detection accuracy and develop a robust approach capable of identifying both known and emerging threats. Our methodology utilized the Electra Modbus dataset as the source domain and was transferred to the Electra S7Comm dataset as the target domain. Experimental results highlight the effectiveness of GAN-based data augmentation in mitigating the scarcity of attack samples, enhancing both the robustness and generalizability of our detection models. Additionally, the application of pre-trained models through transfer learning played a crucial role in achieving superior performance. Our proposed solution can be deployed within railway networks to strengthen security measures, enabling proactive responses to cyber threats, safeguarding critical infrastructure, and ensuring uninterrupted operations., This work is partially supported by the Spanish Ministry of Science and Innovation under contract PID2021-124463OB-IOO and by the Generalitat de Catalunya, Spain under grant 2021-SGR-00326. Finally, the research leading to these results has received funding from the European Union’s Horizon 2020 research and innovation programme under the HORIZON-AG PHOENI2X (101070586), HORIZON-EU HORSE (101096342) and HORIZON-EU VITAMIN-V (101093062) projects., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Rule-based with machine learning IDS for DDoS attack detection in cyber-physical production systems (CPPS)

  • Hussain, Ayaz|||0000-0002-9081-9567
  • Marín Tordera, Eva|||0000-0002-2039-8277
  • Masip Bruin, Xavier|||0000-0002-4755-556X
  • Leligou, Helen C.
Recent advancements in communication technology have transformed the way the industrial system works. This digitalization has improved the way of communication between different actors involved in cyber physical production systems (CPPS), such as users, suppliers, and manufacturers, thus making the whole process transparent. The utilization of emerging new technologies in CPPS can cause vulnerable spots that can be exploited by attackers to launch sophisticated distributed denial of service (DDoS) attacks, hence threatening the availability of the production systems. Existing machine learning based intrusion detection systems (IDS) often rely on unrealistic datasets for training and validation, thus missing the crucial testing phase with real-time scenarios. The results generated by the ML models are based on predictions at each flow level and cannot provide summarized information about malicious entities. To address this limitation, this study proposed an efficient IDS system that uses both rule-based detection and ML-based approaches to detect DDoS attacks damaging the infrastructure of CPPS. For training and validation of the system, we use real-time network traffic extracted from a real industrial scenario, referred to as Farm-to-Fork (F2F) supply chain system. Both, attacks and normal traffic were captured, and bidirectional features were extracted through CIC-FLOWMETER. We make use of 8 ML supervised and unsupervised approaches to detect the malicious flows; and then a rule-based detection mechanism is used to calculate the frequency of the malicious flows and to assign different severity levels based on the computed frequency. The overall results show that supervised models outperform unsupervised approaches and achieve an accuracy 99.97% and TPR 99.96%. Overall, the weighted accuracy when tested and deployed in a real-time scenario is around 98.71%. The results prove that the system works better when considering real-time scenarios and provides comprehensive information about the detected results that can be used to take different mitigation actions., This work was supported in part by European Union’s Horizon Europe (PHOENi2X) under Grant 101070586, in part by the Spanish Ministry of Science and Innovation funded by MCIN/AEI/10.13039/501100011033 under Grant PID2021-124463OB-I00, in part by ERDF a way of making Europe, and in part by the Catalan Government under Contract 2021 SGR 00326., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Exploring image transformations with diffusion models: A survey of applications and implementation code

  • Arellano García, Silvia
  • Otero Calviño, Beatriz|||0000-0002-9194-559X
  • Tous Liesa, Rubén|||0000-0002-1409-5843
Diffusion Models have become increasingly popular in recent years and their applications span a wide range of fields. This survey focuses on the use of diffusion models in computer vision, specially in the branch of image transformations. The objective of this survey is to provide an overview of state-of-the-art applications of diffusion models in image transformations, including image inpainting, super-resolution, restoration, translation, and editing. This survey presents a selection of notable papers and repositories including practical applications of diffusion models for image transformations. The applications are presented in a practical and concise manner, facilitating the understanding of concepts behind diffusion models and how they function. Additionally, it includes a curated collection of GitHub repositories featuring popular examples of these subjects., This work is partially supported by the Spanish Ministry of Science and Innovation under contract PID2019-107255GB and PID2021-124463OB- IOO, by the Generalitat de Catalunya under grants 2021-SGR-00478 and 2021-SGR-00326. Finally, the research leading to these results also has received funding from the European Union’s Horizon 2020 research and innovation programme under the HORIZON-EU VITAMIN-V (101093062) project., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

A survey on security enhancing Digital Twins: models, applications and tools

  • Qureshi, Abdul Rehman|||0009-0005-1797-392X
  • Asensio Garcia, Adrian|||0000-0002-3345-6093
  • Imran, Muhammad|||0000-0001-5109-3700
  • García Almiñana, Jordi|||0000-0002-3515-7150
  • Masip Bruin, Xavier|||0000-0002-4755-556X
A Digital Twin (DT) is a virtual representation of both cyber and physical objects. Initially, the concept was introduced to optimize manufacturing products, aimed at validating and verifying their behavior along their lifecycle. However, in the recent years and with the advent of Industry 4.0, it has also gained huge attention in the cybersecurity arena. Indeed, using a DT is seen as an strategy to automate, monitor, simulate, and diagnose cyber–physical systems behavior, aimed at making them to be secure. The DT can provide security in multiple domains such as manufacturing, aerospace, automotive, construction, smart grid, smart cities, and smart transportation. The fundamentals of a DT is a combination of data and a model of the system. To this end, this survey aims to explore current efforts on DT modeling techniques, and how these DT models capture the behavior of the system to enhance cybersecurity. The study focuses on the construction of behavioral models, techniques, tools, and technologies of DT used for specific security applications. The purpose of the discussion carried out in this survey is to illustrate the reader on how these behavioral models adapt to different security applications. In addition, the paper also analyzes and classifies current DT literature based on use cases, type of DT study, tools, and various DT operation modes with their specific security application. The findings of this survey describe security applications being enhanced by DT, and how the combination of key technologies facilitating Industry 4.0 (i.e., AI, IoT, Big Data and Cloud Computing, among others), are accelerating the adoption of DT as a solution for next generation security applications. Finally, this survey highlights relevant research gaps that need to be addressed before the wide adoption of DT in cybersecurity., The work leading to this publication was partially co-funded by the Spanish Ministry of Science and Innovation under grant PID2021-124463OB-I00 funded by MCIN/AEI/10.13039/501100011033 and by ERDF A way of making Europe, by the Catalan Government under contract 2021 SGR 00326, as well as by the European Union’s Horizon Europe under the HE COBALT project , Grant Agreement no. 101119602., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Deploying vision foundation AI models on the edge. The SAM2 experience

  • Lin, Zheshuo
  • Tous Liesa, Rubén|||0000-0002-1409-5843
  • Otero Calviño, Beatriz|||0000-0002-9194-559X
As AI-driven applications expand across industries, the need for efficient edge computing solutions becomes increasingly critical. Traditional AI models are designed for high-performance cloud infrastructures, but emerging constraints—such as privacy requirements, network limitations, and real-time processing needs—necessitate optimized deployment on resource-constrained edge devices. This study presents a practical experience in adapting Segment Anything Model 2 (SAM2), a vision foundation model, for edge AI environments. The adaptation process involved translating the model to C++ using ONNX Runtime, enabling efficient execution on heterogeneous hardware. Experimental evaluations demonstrate that deploying SAM2 at the edge enhances processing efficiency, reduces reliance on network stability, and improves real-time responsiveness. This research provides valuable insights into AI in pervasive computing environments, contributing to the sustainable and scalable deployment of foundation models on edge devices., This work is partially funded by the Spanish Ministry of Science and Innovation under contracts PID2019-107255GB and PID2021-124463OB-IOO, as well as by the SGR programs 2021-SGR-00478 and 2021-SGR-00326 of the Catalan government. Finally, the research that led to these results has received funding from the Collaboration Grant for the 2024/2025 period from the Spanish Ministry of Education., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Malware detection on Linux using runtime opcode tracing

  • Alonso García, Martí
  • Costa Prats, Juan José|||0000-0003-2479-0230
  • Morancho Llena, Enrique|||0000-0003-2403-8145
The fast-paced evolution of cyberattacks to digital infrastructures requires new protection mechanisms to counter-attack them. Malware attacks, a type of cyberattacks ranging from viruses and worms to ransomware and spyware, have been traditionally detected using signature-based methods. But with new versions of malware, this approach is not good enough, and new machine learning tools look promising. In this paper we evaluate five machine learning models (SVM, KNN, Naive Bayes, Decision Tree and Random Forest) to detect Linux malware on applications during its runtime by tracking their executed instructions (opcodes). We show the methodology, the initial dataset preparation, the infrastructure used to obtain the traces of executed instructions, and the evaluation of the results for the different models used. The obtained results show that Random Forest classifier gets the best results, with 90% accuracy or higher., Funded by the European Union. Views and opinions expressed are, however, those of the authors only and do not necessarily reflect those of the European Union or the HaDEA. Neither the European Union nor the granting authority can be held responsible for them. Project number: 101093062 This work is also partially supported by the Spanish Ministry of Science and Innovation under contract PID2021- 124463OB-IOO and by the Generalitat de Catalunya under grant 2021SGR00326., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

High-resolution image generation using artificial intelligence and diffusion modelling

  • Scoles García, Amanda
  • Sionis Ponte, Giulia
  • Otero Calviño, Beatriz|||0000-0002-9194-559X
  • Utrera Iglesias, Gladys Miriam|||0000-0002-0637-1297
Super-Resolution via Repeated Refinement (SR3) is a state-of-the-art super-resolution algorithm based on diffusion model that can enhance the resolution of images. This method is used to pre-trained models on large datasets and can be used for various tasks without requiring training from scratch. Training SR3 from scratch using the ImageNet dataset involves a complex process that requires substantial computational resources and expertise. The idea is applied the trained SR3 model to new images by feeding the low-resolution inputs and obtaining the high-resolution outputs. It's important to note that training SR3 from scratch is a resource-intensive process that requires powerful GPUs and significant computation time. If you do not have access to such resources, an alternative is to use pre-trained models that are already available and fine-tune them on specific datasets or tasks. The paper shows the result of comparing the resolution of the preprocessed images using a significantly smaller number of images to perform the training with those obtained using the pre-trained model. The results obtained show acceptable results without having to perform on large datasets minimizing the computation time to obtain the resolution of images., This work is partially supported by the Spanish Ministry of Science and Innovation under contract PID2021-124463OB-100 and PID2019-107255GB-C22 by the Generalitat de Catalunya under grant 2021-SGR-00326 and 2021-SGR-01007. Finally, the research leading to these results also has received funding from the European Union's Horizon 2020 research and innovation program under the HORIZON-EU VITAMIN-V (101093062) project., Peer Reviewed
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Rule-based with machine learning IDS for DDoS attack detection in cyber-physical production systems (CPPS)

  • Hussain, Ayaz
  • Marín Tordera, Eva
  • Masip Bruin, Xavier
  • Leligou, Helen C.
Recent advancements in communication technology have transformed the way the industrial system works. This digitalization has improved the way of communication between different actors involved in cyber physical production systems (CPPS), such as users, suppliers, and manufacturers, thus making the whole process transparent. The utilization of emerging new technologies in CPPS can cause vulnerable spots that can be exploited by attackers to launch sophisticated distributed denial of service (DDoS) attacks, hence threatening the availability of the production systems. Existing machine learning based intrusion detection systems (IDS) often rely on unrealistic datasets for training and validation, thus missing the crucial testing phase with real-time scenarios. The results generated by the ML models are based on predictions at each flow level and cannot provide summarized information about malicious entities. To address this limitation, this study proposed an efficient IDS system that uses both rule-based detection and ML-based approaches to detect DDoS attacks damaging the infrastructure of CPPS. For training and validation of the system, we use real-time network traffic extracted from a real industrial scenario, referred to as Farm-to-Fork (F2F) supply chain system. Both, attacks and normal traffic were captured, and bidirectional features were extracted through CIC-FLOWMETER. We make use of 8 ML supervised and unsupervised approaches to detect the malicious flows; and then a rule-based detection mechanism is used to calculate the frequency of the malicious flows and to assign different severity levels based on the computed frequency. The overall results show that supervised models outperform unsupervised approaches and achieve an accuracy 99.97% and TPR 99.96%. Overall, the weighted accuracy when tested and deployed in a real-time scenario is around 98.71%. The results prove that the system works better when considering real-time scenarios and provides comprehensive information about the detected results that can be used to take different mitigation actions., This work was supported in part by European Union’s Horizon Europe (PHOENi2X) under Grant 101070586, in part by the Spanish Ministry of Science and Innovation funded by MCIN/AEI/10.13039/501100011033 under Grant PID2021-124463OB-I00, in part by ERDF a way of making Europe, and in part by the Catalan Government under Contract 2021 SGR 00326., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Low-cost training of image-to-image diffusion models with incremental learning and task/domain adaptation

  • Antona Pizà, Héctor
  • Otero Calviño, Beatriz
  • Tous Liesa, Rubén
Diffusion models specialized in image-to-image translation tasks, like inpainting and colorization, have outperformed the state of the art, yet their computational requirements are exceptionally demanding. This study analyzes different strategies to train image-to-image diffusion models in a low-resource setting. The studied strategies include incremental learning and task/domain transfer learning. First, a base model for human face inpainting is trained from scratch with an incremental learning strategy. The resulting model achieves an FID score almost equivalent to that of its batch learning equivalent while significantly reducing the training time. Second, the base model is fine-tuned to perform a different task, image colorization, and, in a different domain, landscape images. The resulting colorization models showcase exceptional performances with a minimal number of training epochs. We examine the impact of different configurations and provide insights into the ability of image-to-image diffusion models for transfer learning across tasks and domains., This work is partially supported by the Spanish Ministry of Science and Innovation under contracts PID2019-107255GB and PID2021-124463OB-IOO, by the Generalitat de Catalunya under grants 2017-SGR-962, 2021-SGR-00326, 2021-SGR-00478 and DRAC (IU16-011591). The research leading to these results has received funding from the European Union’s Horizon 2020 research and innovation program under the HORIZON-AG PHOENI2X (101070586) and HORIZON-EU VITAMIN-V (101093062) projects., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

A survey on security enhancing Digital Twins: models, applications and tools

  • Qureshi, Abdul Rehman
  • Asensio Garcia, Adrian
  • Imran, Muhammad
  • García Almiñana, Jordi
  • Masip Bruin, Xavier
A Digital Twin (DT) is a virtual representation of both cyber and physical objects. Initially, the concept was introduced to optimize manufacturing products, aimed at validating and verifying their behavior along their lifecycle. However, in the recent years and with the advent of Industry 4.0, it has also gained huge attention in the cybersecurity arena. Indeed, using a DT is seen as an strategy to automate, monitor, simulate, and diagnose cyber–physical systems behavior, aimed at making them to be secure. The DT can provide security in multiple domains such as manufacturing, aerospace, automotive, construction, smart grid, smart cities, and smart transportation. The fundamentals of a DT is a combination of data and a model of the system. To this end, this survey aims to explore current efforts on DT modeling techniques, and how these DT models capture the behavior of the system to enhance cybersecurity. The study focuses on the construction of behavioral models, techniques, tools, and technologies of DT used for specific security applications. The purpose of the discussion carried out in this survey is to illustrate the reader on how these behavioral models adapt to different security applications. In addition, the paper also analyzes and classifies current DT literature based on use cases, type of DT study, tools, and various DT operation modes with their specific security application. The findings of this survey describe security applications being enhanced by DT, and how the combination of key technologies facilitating Industry 4.0 (i.e., AI, IoT, Big Data and Cloud Computing, among others), are accelerating the adoption of DT as a solution for next generation security applications. Finally, this survey highlights relevant research gaps that need to be addressed before the wide adoption of DT in cybersecurity., The work leading to this publication was partially co-funded by the Spanish Ministry of Science and Innovation under grant PID2021-124463OB-I00 funded by MCIN/AEI/10.13039/501100011033 and by ERDF A way of making Europe, by the Catalan Government under contract 2021 SGR 00326, as well as by the European Union’s Horizon Europe under the HE COBALT project , Grant Agreement no. 101119602., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

A machine-learning-based framework for detection and recommendation in response to cyberattacks in critical energy infrastructures

  • Rabadan Arroyo, Raul
  • Hussain, Ayaz
  • Simó Mezquita, Ester
  • Rodríguez Luna, Eva
  • Masip Bruin, Xavier
This paper presents an attack detection, response, and recommendation framework designed to protect the integrity and operational continuity of IoT-based critical infrastructure, specifically focusing on an energy use case. With the growing deployment of IoT-enabled smart meters in energy systems, ensuring data integrity is essential. The proposed framework monitors smart meter data in real time, identifying deviations that may indicate data tampering or device malfunctions. The system comprises two main components: an attack detection and prediction module based on machine learning (ML) models and a response and adaptation module that recommends countermeasures. The detection module employs a forecasting model using a long short-term memory (LSTM) architecture, followed by a dense layer to predict future readings. It also integrates a statistical thresholding technique based on Tukey’s fences to detect abnormal deviations. The system was evaluated on real smart meter data in a testbed environment. It achieved accurate forecasting (MAPE < 2% in most cases) and successfully flagged injected anomalies with a low false positive rate, an effective result given the lightweight, unsupervised, and real-time nature of the approach. These findings confirm the framework’s applicability in resource-constrained energy systems requiring real-time cyberattack detection and mitigation., This work has been partially supported by the HE PHOENI2X project, funded by European Commission with Grant Number 101070586, and by the Spanish Ministry of Science and Innovation under grant PID2021-124463OB-I00, funded by MCIN/AEI/10.13039/501100011033 and by ERDF A way of making Europe, as well as by the Catalan Government under contract 2021 641 SGR 00326., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Automatic Linux malware detection using binary inspection and runtime opcode tracing

  • Alonso García, Martí
  • Gironès, Andreu
  • Costa Prats, Juan José
  • Morancho Llena, Enrique
  • Di Carlo, Stefano
  • Canal Corretger, Ramon
The fast-paced evolution of cyberattacks to digital infrastructures requires new protection mechanisms to counterattack them. Malware attacks, a type of cyberattacks ranging from viruses and worms to ransomware and spyware, have been traditionally detected using signature-based methods. But with new versions of malware, this approach is not good enough, and new machine learning tools look promising. In this paper we present two methods to detect Linux malware using machine learning models: (1) a dynamic approach, that tracks the application executed instructions (opcodes) while they are being executed; and (2) a static approach, that inspects the binary application files before execution. We evaluate (1) five machine learning models (Support Vector Machine, k-Nearest Neighbor, Naive Bayes, Decision Tree and Random Forest) and (2) a deep neural network using a Long Short-Term Memory architecture with word embedding. We show the methodology, the initial dataset preparation, the infrastructure used to obtain the traces of executed instructions, and the evaluation of the results for the different models used. The obtained results show that the dynamic approach with a Random Forest classifier gets a 90% accuracy or higher, while the static approach obtains a 98% accuracy., We thank the anonymous reviewers for their thoughtful comments and suggestions which have improved the presentation of this work. This work is partially supported by the Spanish Ministry of Science and Innovation under contract PID2021-124463OB-IOO and by the Generalitat de Catalunya, Spain under grant 2021-SGR-00326. Finally, the research leading to these results has received funding from the European Union’s Horizon Programme under the HORIZON-EU VITAMIN-V (101093062) project., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Deep-learning based detection for cyber-attacks in IoT networks: A distributed attack detection framework

  • Jullian Parra, Olivia
  • Otero Calviño, Beatriz
  • Rodríguez Luna, Eva
  • Gutiérrez Escobar, Norma
  • Antona Pizà, Héctor
  • Canal Corretger, Ramon
The widespread use of smart devices and the numerous security weaknesses of networks has dramatically increased the number of cyber-attacks in the internet of things (IoT). Detecting and classifying malicious traffic is key to ensure the security of those systems. This paper implements a distributed framework based on deep learning (DL) to prevent many different sources of vulnerability at once, all under the same protection system. Two different DL models are evaluated: feed forward neural network and long short-term memory. The models are evaluated with two different datasets (i.e.NSL-KDD and BoT-IoT) in terms of performance and identification of different kinds of attacks. The results demonstrate that the proposed distributed framework is effective in the detection of several types of cyber-attacks, achieving an accuracy up to 99.95% across the different setups., Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature. This work is partially supported by the Spanish Ministry of Science and Innovation under contract PID2021-124463OB-IOO, by the Generalitat de Catalunya under grants 2017SGR962, 2021SGR00326, and by the DRAC (IU16-011591), the HORIZON Vitamin-V (101093062) and the HORIZON-AG PHOENI2X (101070586) projects., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

A deep learning framework for safety monitoring of a railway section

  • Chriki Zerrouk, Fatima Zohra
  • Simó Mezquita, Ester
  • Aguiló Gost, Francisco de Asis L.
  • Garcia Milà, Ignasi
  • Masip Bruin, Xavier
© 2024 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works., This paper presents an approach for anomaly detection and forecasting that aims to protect the integrity and operational continuity of IoT critical infrastructure, specifically for a railway use case. This system is designed to monitor in real-time the sensor data to detect any deviations that may indicate a potential data tampering attack or a sensor malfunction. This early identification and notification of such anomalies is crucial for preventing unauthorized access and mitigating the risks associated with data tampering attacks. The proposed system is comprises of two primary components: a forecasting component and an anomaly detection component. The Forecasting Component uses a Long Short-Term Memory (LSTM) model to predict future sensor values, while the Anomaly Detection Component employs the Tukey's fence method to identify sensor measurements that significantly deviate from normal behaviour. A railway use case is included to demonstrate the practical application of the deep learning framework. These components were evaluated and both demonstrated excellent performance. The Forecasting Component provided highly accurate predictions of future sensor values, while the Anomaly Detection Component effectively identified deviations from normal patterns. The evaluation results confirmed the system's ability to detect significant anomalies and its capability to maintain operational integrity and security in IoT critical infrastructures., The work leading to this publication was partly funded by the European Union’s Horizon Europe under project PHOENi2X, Grant Agreement no. 101070586. For UPC authors this work is partially co-funded by the Spanish Ministry of Science and Innovation under grant PID2021-124463OBI00 funded by MCIN/AEI/10.13039/501100011033 and by ERDF A way of making Europe, as well as by the Catalan Government under contract 2021 SGR 00326., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

A survey on 5G private and B5G network threats and safeguarding AI-based security mechanisms through the layered analysis

  • Tariq, Saman
  • Rodríguez Luna, Eva
  • Masip Bruin, Xavier
  • Díaz Rodríguez, Rodrigo
  • Martrat, Josep
  • Trakadas, Panagiotis
The fifth-generation (5G) mobile network has shifted the paradigm in connectivity, high-speed data transmission, ultra-low latency, ultra-high throughput, multisense transmission, and ultra-high reliability. Today, industries are increasingly adopting 5G private networks, which also handle sensitive data related to business information, trade secrets, and personal data. Attacks on 5G private networks can potentially result in losing a competitive advantage since different security threats and vulnerabilities progressively target these networks. The unique infrastructure of the 5G network architecture and key enabling technologies exposes them to various vulnerabilities that attackers can target to breach sensitive data, steal information, and disrupt critical systems. Therefore, paying special attention to the security issues of 5G private networks is essential. The advancement of future wireless technology came about because of the different and diverse nature of connected devices, in contrast to the previous generation of mobile networks. The B5G network provides support to open network platforms, open interfaces, and the integration of different key enabling technologies helps manage network services and deploy new services needed for diverse requirements. At the same time, it has increased the attack surface compared to previous-generation networks. Therefore, it is imperative to conduct a review that focuses on addressing and classifying the different emerging threats in the private 5G and B5G networks in a distinctive way. In this paper, we have adopted a layered architecture from an industrial use case of 5G private networks to identify and classify different threats in 5G private networks. The study also characterized and modeled the different threats using information on the type of attack, entry points, and impact of the attack on the architecture layer. Moreover, an analysis of key enablers in 5G private and B5G networks and information on security threats and cyber-attacks is also presented. To accommodate the emerging threats in next-generation wireless technology, we have classified and modeled the different threats in the B5G domain using the common involved layer, which includes perception, network, and application layers in Hexa-x E2E and 6G IoT-enabled architecture. Organizations and projects that do not actively engage in cutting-edge technologies will lose their competitive advantage in the evolving technological landscape. This study has mapped the identified threat categories with different types of threats that could occur at different layers and assessed the entry points of the attacker. We have also identified the attack’s impact at each layer using security requirements related to confidentiality, availability, and integrity. Furthermore, this study reviews different AI-enabled solutions that can add value in preserving the security of 5G and B5G networks. Through this review analysis, we can conclude that the layered approach is quite beneficial in identifying the different threats and what security solutions can be offered to mitigate them., The work leading to this publication was partly funded by the European Union’s Horizon Europe under the project HORSE, Grant Agreement no. 101096342. For UPC authors this work is partially cofunded by the Spanish Ministry of Science and Innovation under grant PID2021-124463OB-I00 funded by MCIN/AEI/10.13039/501100011033 and by ERDF A way of making Europe, as well as by the Catalan Government under contract 2021 SGR 00326., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

A comprehensive architecture for federated learning-based smart advertising

  • Seyghaly, Rasool
  • García Almiñana, Jordi
  • Masip Bruin, Xavier
This paper introduces a cutting-edge data architecture designed for a smart advertising context, prioritizing efficient data flow and performance, robust security, while guaranteeing data privacy and integrity. At the core of this study lies the application of federated learning (FL) as the primary methodology, which emphasizes the authenticity and privacy of data while promptly discarding irrelevant or fraudulent information. Our innovative data model employs a semi-random role assignment strategy based on a variety of criteria to efficiently collect and amalgamate data. The architecture is composed of model nodes, data nodes, and validator nodes, where the role of each node is determined by factors such as computational capability, interconnection quality, and historical performance records. A key feature of our proposed system is the selective engagement of a subset of nodes for modeling and validation, optimizing resource use and minimizing data loss. The AROUND social network platform serves as a real-world case study, illustrating the efficacy of our data architecture in a practical setting. Both simulated and real implementations of our architecture showcase its potential to dramatically curtail network traffic and average CPU usage, while preserving the accuracy of the FL model. Remarkably, the system is capable of achieving over a 50% reduction in both network traffic and average CPU usage even when the user count escalates by twenty-fold. The click rate, user engagement, and other parameters have also been evaluated, proving that the proposed architecture’s advantages do not affect the smart advertising accuracy. These findings highlight the proposed architecture’s capacity to scale efficiently and maintain high performance in smart advertising environments, making it a valuable contribution to the evolving landscape of digital marketing and FL., This work has been supported by the Spanish Ministry of Science and Innovation under grant PID2021-124463OB-I00 funded by MCIN/AEI/10.13039/501100011033 and by ERDF A way of making Europe, and by the Catalan Government under contract 2021 SGR 00326. The corresponding author, R.S., gratefully acknowledges the Universitat Politècnica de Catalunya and Banco Santander for the financial support of his predoctoral grant FPI-UPC 2021, Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

High-resolution image generation using artificial intelligence and diffusion modelling

  • Scoles García, Amanda
  • Sionis Ponte, Giulia
  • Otero Calviño, Beatriz
  • Utrera Iglesias, Gladys Miriam
Super-Resolution via Repeated Refinement (SR3) is a state-of-the-art super-resolution algorithm based on diffusion model that can enhance the resolution of images. This method is used to pre-trained models on large datasets and can be used for various tasks without requiring training from scratch. Training SR3 from scratch using the ImageNet dataset involves a complex process that requires substantial computational resources and expertise. The idea is applied the trained SR3 model to new images by feeding the low-resolution inputs and obtaining the high-resolution outputs. It's important to note that training SR3 from scratch is a resource-intensive process that requires powerful GPUs and significant computation time. If you do not have access to such resources, an alternative is to use pre-trained models that are already available and fine-tune them on specific datasets or tasks. The paper shows the result of comparing the resolution of the preprocessed images using a significantly smaller number of images to perform the training with those obtained using the pre-trained model. The results obtained show acceptable results without having to perform on large datasets minimizing the computation time to obtain the resolution of images., This work is partially supported by the Spanish Ministry of Science and Innovation under contract PID2021-124463OB-100 and PID2019-107255GB-C22 by the Generalitat de Catalunya under grant 2021-SGR-00326 and 2021-SGR-01007. Finally, the research leading to these results also has received funding from the European Union's Horizon 2020 research and innovation program under the HORIZON-EU VITAMIN-V (101093062) project., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Innovative predictive approach towards a personalized oxygen dosing system

  • Pascual Saldaña, Heribert
  • Masip Bruin, Xavier
  • Asensio Garcia, Adrian
  • Alonso Beltran, Albert
  • Blanco Vich, Isabel
Despite the large impact chronic obstructive pulmonary disease (COPD) that has on the population, the implementation of new technologies for diagnosis and treatment remains limited. Current practices in ambulatory oxygen therapy used in COPD rely on fixed doses overlooking the diverse activities which patients engage in. To address this challenge, we propose a software architecture aimed at delivering patient-personalized edge-based artificial intelligence (AI)-assisted models that are built upon data collected from patients’ previous experiences along with an evaluation function. The main objectives reside in proactively administering precise oxygen dosages in real time to the patient (the edge), leveraging individual patient data, previous experiences, and actual activity levels, thereby representing a substantial advancement over conventional oxygen dosing. Through a pilot test using vital sign data from a cohort of five patients, the limitations of a one-size-fits-all approach are demonstrated, thus highlighting the need for personalized treatment strategies. This study underscores the importance of adopting advanced technological approaches for ambulatory oxygen therapy., This research was funded by the Spanish Ministry of Science, Innovation and Universities and FEDER, grant number PID2021-124463OB-100, and by the AGAUR Catalan Agency, grant number 2021_SGR_00326., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Exploring image transformations with diffusion models: A survey of applications and implementation code

  • Arellano García, Silvia
  • Otero Calviño, Beatriz
  • Tous Liesa, Rubén
Diffusion Models have become increasingly popular in recent years and their applications span a wide range of fields. This survey focuses on the use of diffusion models in computer vision, specially in the branch of image transformations. The objective of this survey is to provide an overview of state-of-the-art applications of diffusion models in image transformations, including image inpainting, super-resolution, restoration, translation, and editing. This survey presents a selection of notable papers and repositories including practical applications of diffusion models for image transformations. The applications are presented in a practical and concise manner, facilitating the understanding of concepts behind diffusion models and how they function. Additionally, it includes a curated collection of GitHub repositories featuring popular examples of these subjects., This work is partially supported by the Spanish Ministry of Science and Innovation under contract PID2019-107255GB and PID2021-124463OB- IOO, by the Generalitat de Catalunya under grants 2021-SGR-00478 and 2021-SGR-00326. Finally, the research leading to these results also has received funding from the European Union’s Horizon 2020 research and innovation programme under the HORIZON-EU VITAMIN-V (101093062) project., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

SBNNR: small-size bat-optimized KNN regression

  • Seyghaly, Rasool
  • García Almiñana, Jordi
  • Masip Bruin, Xavier
  • Kuljanin, Jovana
Small datasets are frequent in some scientific fields. Such datasets are usually created due to the difficulty or cost of producing laboratory and experimental data. On the other hand, researchers are interested in using machine learning methods to analyze this scale of data. For this reason, in some cases, low-performance, overfitting models are developed for small-scale data. As a result, it appears necessary to develop methods for dealing with this type of data. In this research, we provide a new and innovative framework for regression problems with a small sample size. The base of our proposed method is the K-nearest neighbors (KNN) algorithm. For feature selection, instance selection, and hyperparameter tuning, we use the bat optimization algorithm (BA). Generative Adversarial Networks (GANs) are employed to generate synthetic data, effectively addressing the challenges associated with data sparsity. Concurrently, Deep Neural Networks (DNNs), as a deep learning approach, are utilized for feature extraction from both synthetic and real datasets. This hybrid framework integrates KNN, DNN, and GAN as foundational components and is optimized in multiple aspects (features, instances, and hyperparameters) using BA. The outcomes exhibit an enhancement of up to 5% in the coefficient of determination (¿^2 score) using the proposed method compared to the standard KNN method optimized through grid search., This work has been supported by the Spanish Ministry of Science and Innovation under grant PID2021-124463OB-I00, funded by MCIN/AEI/10.13039/501100011033 and by ERDF “A way of making Europe”, by the European Union’s Horizon Europe under the HE ICOS project, Grant Agreement no. 101070177, and by the Catalan Government under contract 2021 SGR 00326. The corresponding author R.S. gratefully acknowledges the Universitat Politècnica de Catalunya and Banco Santander for the financial support of his predoctoral grant FPI-UPC 2021., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

An NIDS for known and zero-day anomalies

  • Hussain, Ayaz
  • Aguiló Gost, Francisco de Asis L.
  • Simó Mezquita, Ester
  • Marín Tordera, Eva
  • Masip Bruin, Xavier
Rapid development in the network infrastructure has resulted in sophisticated attacks which are hard to detect using typical network intrusion detection systems (NIDS). There is a strong need for efficient NIDS to detect these known attacks along with ever-emerging zero-day exploits. Existing NIDS are more focused on detecting known attacks using supervised machine learning approaches, achieving better performance for known attacks but poor detection of unknown attacks. Many NIDS have utilized the unsupervised approach, which results in better detection of unknown anomalies. In this paper, we proposed a Hybrid NIDS based on Semisupervised One-Class Support Vector Machine (OC-SVM) and Supervised Random Forest (RF) algorithms. This detection system has several stages. The First stage is based on OC-SVM, which filters benign and malicious traffic. The next stages use many parallel supervised models and an additional OC-SVM model to separate known and unknown attacks from malicious traffic. The previous process is done so that known attacks are classified by their type, and unknown attacks are detected. The proposed NIDS is tested on the standard public dataset CSE-CIC-IDS-2018. The evaluation results show that the system achieves a high accuracy, 99.45%, for detecting known attacks. Our proposed NIDS achieves an accuracy of 93.99% for unknown or zero-day attacks. The overall accuracy of the proposed NIDS is 95.95%. The system significantly improves the detection of known and unknown anomalies using a hybrid approach., This project has received funding from the European Union’s H2020 research and innovation programme under the grant agreement No. 952644, from the Spanish Ministry of Science and Innovation under contract PID2021-124463OB- I00, and from the Catalan Government under contract 2021 SGR 00326., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Use of automatic 6-minute walking test recording system in patients with chronic respiratory diseases

  • Torres Castro, Rodrigo
  • Pascual Saldaña, Heribert
  • Alonso Beltran, Albert
  • Gimeno Santos, Elena
  • Palomo Gil, Maria
  • Barberà Mir, Joan Albert
  • Masip Bruin, Xavier
  • Batlle, Jaume
  • Bigorra Llosas, Joan
  • Blanco Vich, Isabel
To the editor: Physical capacity tests are used for the diagnosis, prognosis and monitoring of chronic respiratory diseases (CRD), such as Chronic Obstructive Pulmonary Disease (COPD) or Pulmonary Hypertension (PH).1 Among these tests, the most widely used is the 6-minute walking test (6MWT).2 Additionally, the guidelines highlight the need to continuously record biological signals, particularly oxygen saturation (SpO2), which has been shown to be a prognostic marker in CRD.2 [...], The study was supported by grants from Societat Catalana de Pneumologia (SOCAP) (ESTEVE TEIJIN/2021), PI17/1515 and PI21/0555 from the Instituto de Salud Carlos III (ISCiii), co-funded by the European Union (ERDF/ESF, ”A way to make Europe” “Investing in your future”) and Premi d’Innovació de l’Hospital Clínic de Barcelona (2021). For UPC authors this work has been supported by the Spanish Ministry of Science and Innovation under contract PID2021-124463OB-I00, the Catalan Government under contract 2021 SGR 00326 and the Catalan Department of Research and Universities., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Analyzing distinct neural network models for oxygen saturation prediction towards a personalized COPD management

  • Pascual Saldaña, Heribert
  • Masip Bruin, Xavier
  • Alonso Beltran, Albert
  • Blanco Vich, Isabel
Chronic Respiratory Diseases (CRDs), including Chronic Obstructive Pulmonary Disease (COPD), pose significant global health challenges. Long-term oxygen therapy (LTOT) has been widely adopted to improve respiratory function and quality of life in CRD patients. However, current oxygen dosing approaches show several limitations, including inconvenience and mismatched oxygen flow rates according to patients' activity. To address these challenges, the design of Patient Specific Modeling (PSM) and the utilization of advanced technologies, such as artificial intelligence (AI), can be integrated towards a personalized oxygen dosing. This study presents a first draft of a conceptual architecture combining cloud and edge computing aimed at optimizing oxygen dosing based on individual patient data. The evaluation of sixty neural network architectures in a small-scale test has shown good results, with smaller networks generally outperforming larger ones. Further research is needed to design new testing procedures and explore optimal parameters for the neural networks, including a large patients' number., This work has been supported by the Spanish Ministry of Science and Innovation under grant PID2021-124463OB-I00 funded by MCIN/AEI/10.13039/501100011033 and by ERDF A way of making Europe, the Catalan Government under contract 2021 SGR 00326 and the Catalan Department of Research and Universities, Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

A security services management architecture toward resilient 6G wireless and computing ecosystems

  • Rodríguez Luna, Eva
  • Masip Bruin, Xavier
  • Martrat, Josep
  • Díaz, Rodrigo
  • Jukan, Admela
  • Granelli, Fabrizio
  • Trakadas, Panagiotis
  • Xilouris, George
The upcoming sixth-generation (6G) networks are expected to be more heterogeneous, scalable, reliable, secure and energy-efficient. 6G services and applications will benefit from the fast data transmission rates, global coverage, precise positioning, and widespread intelligence capabilities provided by 6G technology. In this complex scenario, new attack surfaces and vectors will emerge, due to the distributed, disaggregated, dynamic, programmable and open nature of the entire end-to-end 6G infrastructure along with the fragmentation of data, as well as the need for supporting cross-platforms interoperability, making the application of security and trust fundamentally challenging. To securely manage services in the future 6G mobile networks along with the set of functions ruling its management, conceptual work is needed to defining functional blocks supporting a secure end-to-end 6G systems management. This is especially critical in handling specific functionalities, such as network disaggregation, risk and threats business impact, energyefficiency, self-configuration or dynamic discovery. This paper positions an innovative 6G security services management architecture, which builds on a set of innovative building blocks, leveraging key technologies, such as AI-assisted models and Digital Twin, while enabling a human-centric approach toward an end-to-end security solution. Two canonical workflows along with two illustrative application scenarios are proposed, Light Rail Transit and Extended Reality, to conceptually highlight the proposed architecture impact and its expected benefits when high bandwidth, ultra-low latency, and secure communications are required., This work was supported by the European union’s Horizon Europe, under the project HORSE, Grant 101096342. The work of Eva Rodriguez and Xavi Masip-Bruin was supported in part by the Spanish Ministry of Science and Innovation funded by MCIN/AEI/10.13039/501100011033 and ERDF: A way of making Europe under Grant PID2021-124463OB-I00, and in part by the Catalan Government under Contract 2021 SGR 00326., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

A 3D terrain generator: Enhancing robotics simulations with GANs

  • Arellano García, Silvia
  • Otero Calviño, Beatriz
  • Kucner, Tomasz Piotr
  • Canal Corretger, Ramon
Simulation is essential in robotics to evaluate models and techniques in a controlled setting before conducting experiments on tangible agents. However, developing simulation environments can be a challenging and time-consuming task. To address this issue, a proposed solution involves building a functional pipeline that generates 3D realistic terrains using Generative Adversarial Networks (GANs). By using GANs to create terrain, the pipeline can quickly and efficiently generate detailed surfaces, saving researchers time and effort in developing simulation environments for their experiments. The proposed model utilizes a Deep Convolutional Generative Adversarial Network (DCGAN) to generate heightmaps, which are trained on a custom database consisting of real heightmaps. Furthermore, an Enhanced Super-Resolution Generative Adversarial Network (ESRGAN) is used to improve the resolution of the resulting heightmaps, enhancing their visual quality and realism. To generate a texture according to the topography of the heightmap, chroma keying is used with previously selected textures. The heightmap and texture are then rendered and integrated, resulting in a realistic 3D terrain. Together, these techniques enable the model to generate high-quality, realistic 3D terrains for use in robotic simulators, allowing for more accurate and effective evaluations of robotics models and techniques., This work is partially supported by the Spanish Ministry of Science and Innovation under contract PID2021-124463OB-IOO, by the Generalitat de Catalunya under grant 2021-SGR-00326. Finally, the research leading to these results also has received funding from the European Union’s Horizon 2020 research and innovation programme under the HORIZON-EU VITAMIN-V (101093062) project., Peer Reviewed, Objectius de Desenvolupament Sostenible::9 - Indústria, Innovació i Infraestructura, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Enhancing cybersecurity in railways: Machine learning approaches for attack detection

  • Otero Calviño, Beatriz
  • Rodríguez Luna, Eva
  • Costa Prats, Juan José
  • Oriol, Mercedes
Ensuring the security of railway systems is crucial to protecting both passengers and infrastructure from the increasing threat of cyberattacks. As these threats grow in complexity and frequency, the need for resilient attack detection systems becomes more pressing. This study tackles the challenge of attack detection in railway systems using machine learning techniques. By integrating Generative Adversarial Networks (GANs), Convolutional Neural Networks (CNNs), and Transfer Learning (TL), we enhance detection accuracy and develop a robust approach capable of identifying both known and emerging threats. Our methodology utilized the Electra Modbus dataset as the source domain and was transferred to the Electra S7Comm dataset as the target domain. Experimental results highlight the effectiveness of GAN-based data augmentation in mitigating the scarcity of attack samples, enhancing both the robustness and generalizability of our detection models. Additionally, the application of pre-trained models through transfer learning played a crucial role in achieving superior performance. Our proposed solution can be deployed within railway networks to strengthen security measures, enabling proactive responses to cyber threats, safeguarding critical infrastructure, and ensuring uninterrupted operations., This work is partially supported by the Spanish Ministry of Science and Innovation under contract PID2021-124463OB-IOO and by the Generalitat de Catalunya, Spain under grant 2021-SGR-00326. Finally, the research leading to these results has received funding from the European Union’s Horizon 2020 research and innovation programme under the HORIZON-AG PHOENI2X (101070586), HORIZON-EU HORSE (101096342) and HORIZON-EU VITAMIN-V (101093062) projects., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Service placement in the continuum: A systematic literature review

  • Sajjad, Waseem
  • Farreras Esclusa, Montserrat
  • García Almiñana, Jordi
  • Masip Bruin, Xavier
Cloud computing plays a crucial role in the Industry 4.0 era, particularly with the rise of Internet of Things (IoT) applications that support domains such as education, healthcare, business, and manufacturing. These applications consist of multiple services with diverse quality of service (QoS) requirements, making their development and deployment complex. While traditional cloud environments provide scalability, they often fail to support latency-sensitive and resource-intensive applications. To overcome these limitations, alternative paradigms such as Cloud–Fog–Edge (CFE), Cloud–Fog (CF), Cloud–Edge (CE), Fog–Edge (FE), and Mobile Edge Computing (MEC) have emerged. These models push computation, storage, and networking closer to end devices, reducing latency and bandwidth usage. However, the heterogeneity, mobility, and dynamic nature of these environments make service placement (a known NP-hard problem) a central challenge. This article presents a systematic literature review of service placement approaches across the compute continuum. Following established SLR methodology, we identified and analyzed 124 peer-reviewed studies published between 2018 and 2024, classifying them by (i) deployment environment, (ii) service placement strategies and algorithms, (iii) adaptability of the solution, (iv) optimization objectives, (v) virtualization/orchestration technologies, (vi) evaluation methodologies, including workloads, testbeds, and simulation tools and (vii) use cases or application types. The novelty of this work lies in providing not only a detailed taxonomy of placement approaches but also this is the first survey that takes all seven aspects into consideration and establishes correlations between them. Our findings reveal that most existing works target smart health applications and favor heuristic-based placement in complex CFE scenarios, while research on scientific and compute-intensive workloads remains limited. We also identify Kubernetes as the most widely used orchestration technology and latency as the dominant optimization metric. Despite significant progress, the field is still maturing, with gaps in real-world validation and adaptive, ML-based placement strategies. By consolidating technical approaches, evaluation practices, and open challenges, this survey offers both researchers and practitioners a structured overview of the state of the art and guidance for advancing service placement in the compute continuum., The work leading to this publication was partially co-funded by the Spanish Ministry of Science and Innovation under grant PID2021-124463OB-I00 funded by MCIN /AEI /10.13039/501100011033 and by the ERDF ‘‘A way of making Europe’’; by the Catalan Government under contract 2021 SGR 00326; and by the EU Horizon programme under grant 101070177 (ICOS project)., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Deploying vision foundation AI models on the edge. The SAM2 experience

  • Lin, Zheshuo
  • Tous Liesa, Rubén
  • Otero Calviño, Beatriz
As AI-driven applications expand across industries, the need for efficient edge computing solutions becomes increasingly critical. Traditional AI models are designed for high-performance cloud infrastructures, but emerging constraints—such as privacy requirements, network limitations, and real-time processing needs—necessitate optimized deployment on resource-constrained edge devices. This study presents a practical experience in adapting Segment Anything Model 2 (SAM2), a vision foundation model, for edge AI environments. The adaptation process involved translating the model to C++ using ONNX Runtime, enabling efficient execution on heterogeneous hardware. Experimental evaluations demonstrate that deploying SAM2 at the edge enhances processing efficiency, reduces reliance on network stability, and improves real-time responsiveness. This research provides valuable insights into AI in pervasive computing environments, contributing to the sustainable and scalable deployment of foundation models on edge devices., This work is partially funded by the Spanish Ministry of Science and Innovation under contracts PID2019-107255GB and PID2021-124463OB-IOO, as well as by the SGR programs 2021-SGR-00478 and 2021-SGR-00326 of the Catalan government. Finally, the research that led to these results has received funding from the Collaboration Grant for the 2024/2025 period from the Spanish Ministry of Education., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Malware detection on Linux using runtime opcode tracing

  • Alonso García, Martí
  • Costa Prats, Juan José
  • Morancho Llena, Enrique
The fast-paced evolution of cyberattacks to digital infrastructures requires new protection mechanisms to counter-attack them. Malware attacks, a type of cyberattacks ranging from viruses and worms to ransomware and spyware, have been traditionally detected using signature-based methods. But with new versions of malware, this approach is not good enough, and new machine learning tools look promising. In this paper we evaluate five machine learning models (SVM, KNN, Naive Bayes, Decision Tree and Random Forest) to detect Linux malware on applications during its runtime by tracking their executed instructions (opcodes). We show the methodology, the initial dataset preparation, the infrastructure used to obtain the traces of executed instructions, and the evaluation of the results for the different models used. The obtained results show that Random Forest classifier gets the best results, with 90% accuracy or higher., Funded by the European Union. Views and opinions expressed are, however, those of the authors only and do not necessarily reflect those of the European Union or the HaDEA. Neither the European Union nor the granting authority can be held responsible for them. Project number: 101093062 This work is also partially supported by the Spanish Ministry of Science and Innovation under contract PID2021- 124463OB-IOO and by the Generalitat de Catalunya under grant 2021SGR00326., Peer Reviewed, Postprint (author's final draft)