{"id":105,"date":"2025-01-14T10:24:08","date_gmt":"2025-01-14T09:24:08","guid":{"rendered":"http:\/\/si2dev2"},"modified":"2026-02-20T14:27:07","modified_gmt":"2026-02-20T13:27:07","slug":"publicaciones","status":"publish","type":"page","link":"https:\/\/incibe.dasci.es\/en\/areas\/investigacion-en-ia-aplicada-a-ciberseguridad\/publicaciones\/","title":{"rendered":"Publications"},"content":{"rendered":"\n<div class=\"wp-block-media-text is-stacked-on-mobile\" style=\"grid-template-columns:40% auto\"><figure class=\"wp-block-media-text__media\"><img fetchpriority=\"high\" decoding=\"async\" width=\"320\" height=\"543\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/10611769-fig-1-source-small.gif\" alt=\"\" class=\"wp-image-9658 size-full\"\/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo &#8211; Defense Strategy against Byzantine Attacks in Federated Machine Learning: Developments towards Explainability<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores<\/strong> &#8211; <a href=\"https:\/\/ieeexplore.ieee.org\/author\/37087027006\">Nuria Rodr\u00edguez-Barroso<\/a>,&nbsp;<a href=\"https:\/\/ieeexplore.ieee.org\/author\/37391702200\">Javier Del Ser<\/a>,&nbsp;<a href=\"https:\/\/ieeexplore.ieee.org\/author\/38196461900\">M. Victoria Luz\u00f3n<\/a>,&nbsp;<a href=\"https:\/\/ieeexplore.ieee.org\/author\/257563195063914\">Francisco Herrera<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen<\/strong> &#8211; The rise of high-risk AI systems has led to escalating concerns, prompting regulatory efforts such as the recently approved EU AI Act. In this context, the development of responsible AI systems is crucial. To this end, trustworthy AI techniques aim at requirements (including transparency, privacy awareness and fairness) that contribute to the development of responsible, robust and safe AI systems. Among them, Federated Learning (FL) has emerged as a key approach to safeguarding data privacy while enabling the collaborative training of AI models. However, FL is prone to adversarial attacks, particularly byzantine attacks, which aim to modify the behavior of the model. This work addresses this issue by proposing an eXplainable and Impartial Dynamic Defense against Byzantine Attacks (XI-DDaBA). This defense mechanism relies on robust aggregation operators and filtering techniques to mitigate the effects of adversarial attacks in FL, while providing explanations for its decisions and ensuring that clients with poor data quality are not discriminated. Experimental simulations are discussed to assess the performance of XI-DDaBA against other baselines from the literature, and to showcase its provided explanations. Overall, XI-DDaBA aligns with the need for responsible AI systems in high-risk collaborative learning scenarios through the explainable and impartial provision of robustness against attacks.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en:&nbsp;<\/strong><a href=\"https:\/\/ieeexplore.ieee.org\/xpl\/conhome\/10609967\/proceeding\">2024 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE)<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI:&nbsp;<\/strong><a href=\"https:\/\/doi.org\/10.1109\/FUZZ-IEEE60900.2024.10611769\" target=\"_blank\" rel=\"noreferrer noopener\">10.1109\/FUZZ-IEEE60900.2024.10611769<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/ieeexplore.ieee.org\/document\/10611769\/authors#authors\">Enlace al art\u00edculo<\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<!--nextpage-->\n\n\n\n<div class=\"wp-block-media-text is-stacked-on-mobile\" style=\"grid-template-columns:40% auto\"><figure class=\"wp-block-media-text__media\"><img decoding=\"async\" width=\"1024\" height=\"817\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/1-s2.0-S1566253524005700-gr1_lrg-1024x817.jpg\" alt=\"\" class=\"wp-image-9663 size-full\" srcset=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/1-s2.0-S1566253524005700-gr1_lrg-1024x817.jpg 1024w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/1-s2.0-S1566253524005700-gr1_lrg-300x239.jpg 300w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/1-s2.0-S1566253524005700-gr1_lrg-768x613.jpg 768w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/1-s2.0-S1566253524005700-gr1_lrg-1536x1226.jpg 1536w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/1-s2.0-S1566253524005700-gr1_lrg-15x12.jpg 15w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/1-s2.0-S1566253524005700-gr1_lrg.jpg 1584w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo &#8211; FLEX: Flexible Federated Learning Framework<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores<\/strong> &#8211; <a href=\"https:\/\/www.sciencedirect.com\/author\/7102347190\/francisco-p-herrera\">F.&nbsp;Herrera<\/a>,&nbsp;D.&nbsp;Jim\u00e9nez-L\u00f3pez,&nbsp;A.&nbsp;Argente-Garrido,&nbsp;N.&nbsp;Rodr\u00edguez-Barroso,&nbsp;C.&nbsp;Zuheros,&nbsp;I.&nbsp;Aguilera-Martos,&nbsp;B.&nbsp;Bello,&nbsp;M.&nbsp;Garc\u00eda-M\u00e1rquez,&nbsp;M.V.&nbsp;Luz\u00f3n<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen<\/strong> &#8211; In the realm of&nbsp;<a href=\"https:\/\/www.sciencedirect.com\/topics\/computer-science\/artificial-intelligence\">Artificial Intelligence<\/a>&nbsp;(AI), the need for privacy and security in data processing has become paramount. As&nbsp;<a href=\"https:\/\/www.sciencedirect.com\/topics\/computer-science\/artificial-intelligence-applications\">AI applications<\/a>&nbsp;continue to expand, the collection and handling of sensitive data raise concerns about individual privacy protection.&nbsp;<a href=\"https:\/\/www.sciencedirect.com\/topics\/computer-science\/federated-learning\">Federated Learning<\/a>&nbsp;(FL) emerges as a promising solution to address these challenges by enabling decentralized model training on local devices, thus preserving data privacy. This paper introduces FLEX: a FLEXible Federated Learning Framework designed to provide maximum flexibility in FL research experiments and the possibility to deploy federated solutions. By offering customizable features for&nbsp;<a href=\"https:\/\/www.sciencedirect.com\/topics\/computer-science\/data-distribution\">data distribution<\/a>, privacy parameters, and communication strategies, FLEX empowers researchers to innovate and develop novel FL techniques. It also provides a distributed version that allows experiments to be deployed on different devices. The framework also includes libraries for specific FL implementations including: (1) anomalies, (2)&nbsp;<a href=\"https:\/\/www.sciencedirect.com\/topics\/computer-science\/blockchain\">blockchain<\/a>, (3)&nbsp;<a href=\"https:\/\/www.sciencedirect.com\/topics\/computer-science\/adversarial-machine-learning\">adversarial attacks<\/a>&nbsp;and defenses, (4)&nbsp;<a href=\"https:\/\/www.sciencedirect.com\/topics\/computer-science\/natural-language-processing\">natural language processing<\/a>&nbsp;and (5)&nbsp;<a href=\"https:\/\/www.sciencedirect.com\/topics\/computer-science\/decision-trees\">decision trees<\/a>, enhancing its versatility and applicability in various domains. Overall, FLEX represents a significant advancement in FL research and deployment, facilitating the development of robust and efficient FL applications.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en:&nbsp;<\/strong><a href=\"https:\/\/www.sciencedirect.com\/journal\/information-fusion\">Information Fusion<\/a> <a href=\"https:\/\/www.sciencedirect.com\/journal\/information-fusion\/vol\/117\/suppl\/C\">Volume 117<\/a>,&nbsp;May 2025, 102792<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI:&nbsp;<\/strong><a href=\"https:\/\/doi.org\/10.1016\/j.inffus.2024.102792\">https:\/\/doi.org\/10.1016\/j.inffus.2024.102792<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/www.sciencedirect.com\/science\/article\/pii\/S1566253524005700#fig1\">Enlace al art\u00edculo<\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<!--nextpage-->\n\n\n\n<div class=\"wp-block-media-text is-stacked-on-mobile\" style=\"grid-template-columns:40% auto\"><figure class=\"wp-block-media-text__media\"><img decoding=\"async\" width=\"1024\" height=\"577\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/make-07-00043-ag-1024x577.webp\" alt=\"\" class=\"wp-image-9667 size-full\" srcset=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/make-07-00043-ag-1024x577.webp 1024w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/make-07-00043-ag-300x169.webp 300w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/make-07-00043-ag-768x433.webp 768w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/make-07-00043-ag-18x10.webp 18w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/make-07-00043-ag.webp 1107w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo &#8211; Membership Inference Attacks Fueled by Few-Shot Learning to Detect Privacy Leakage and Address Data Integrity<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores<\/strong> &#8211; <a href=\"https:\/\/sciprofiles.com\/profile\/author\/ZzNWR1BOcWwyS3hiSWp2YnNDaERKdz09?utm_source=mdpi.com&amp;utm_medium=website&amp;utm_campaign=avatar_name\">Daniel Jim\u00e9nez-L\u00f3pez<\/a>, Nuria Rodr\u00edguez-Barroso, <a href=\"mailto:rbnuria@ugr.es\"><\/a>M. Victoria Luz\u00f3n, Javier Del Ser, <a href=\"mailto:javier.delser@tecnalia.com\"><\/a>Francisco Herrera<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen<\/strong> &#8211; Deep learning models have an intrinsic privacy issue as they memorize parts of their training data, creating a privacy leakage. Membership inference attacks (MIAs) exploit this to obtain confidential information about the data used for training, aiming to steal information. They can be repurposed as a measurement of data integrity by inferring whether the data were used to train a machine learning model. While state-of-the-art attacks achieve significant privacy leakage, their requirements render them infeasible, hindering their use as practical tools to assess the magnitude of the privacy risk. Moreover, the most appropriate evaluation metric of MIA, the true positive rate at a low false positive rate, lacks interpretability. We claim that the incorporation of few-shot learning techniques into the MIA field and a suitable qualitative and quantitative privacy evaluation measure should resolve these issues. In this context, our proposal is twofold. We propose a few-shot learning-based MIA, termed the FeS-MIA model, which eases the evaluation of the privacy breach of a deep learning model by significantly reducing the number of resources required for this purpose. Furthermore, we propose an interpretable quantitative and qualitative measure of privacy, referred to as the Log-MIA measure. Jointly, these proposals provide new tools to assess privacy leakages and to ease the evaluation of the training data integrity of deep learning models, i.e., to analyze the privacy breach of a deep learning model. Experiments carried out with MIA over image classification and language modeling tasks, and a comparison to the state of the art, show that our proposals excel in identifying privacy leakages in a deep learning model with little extra information.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en:&nbsp;<\/strong><em>Mach. Learn. Knowl. Extr.<\/em>&nbsp;<strong>2025<\/strong>,&nbsp;<em>7<\/em>(2), 43<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI:&nbsp;<\/strong><a href=\"https:\/\/doi.org\/10.3390\/make7020043\">https:\/\/doi.org\/10.3390\/make7020043<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/www.mdpi.com\/2504-4990\/7\/2\/43#\">Enlace al art\u00edculo<\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<!--nextpage-->\n\n\n\n<div class=\"wp-block-media-text is-stacked-on-mobile\" style=\"grid-template-columns:40% auto\"><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"641\" height=\"242\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-115312.png\" alt=\"\" class=\"wp-image-9671 size-full\" srcset=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-115312.png 641w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-115312-300x113.png 300w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-115312-18x7.png 18w\" sizes=\"(max-width: 641px) 100vw, 641px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo &#8211; Improving (\u03b1, f)-Byzantine Resilience in Federated Learning via layerwise aggregation and cosine distance<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores<\/strong> &#8211; <a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Garc%C3%ADa-M%C3%A1rquez,+M\">Mario Garc\u00eda-M\u00e1rquez<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Rodr%C3%ADguez-Barroso,+N\">Nuria Rodr\u00edguez-Barroso<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Luz%C3%B3n,+M\">M.Victoria Luz\u00f3n<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Herrera,+F\">Francisco Herrera<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen<\/strong> &#8211; The rapid development of artificial intelligence systems has amplified societal concerns regarding their usage, necessitating regulatory frameworks that encompass data privacy. Federated Learning (FL) is posed as potential solution to data privacy challenges in distributed machine learning by enabling collaborative model training {without data sharing}. However, FL systems remain vulnerable to Byzantine attacks, where malicious nodes contribute corrupted model updates. While Byzantine Resilient operators have emerged as a widely adopted robust aggregation algorithm to mitigate these attacks, its efficacy diminishes significantly in high-dimensional parameter spaces, sometimes leading to poor performing models. This paper introduces Layerwise Cosine Aggregation, a novel aggregation scheme designed to enhance robustness of these rules in such high-dimensional settings while preserving computational efficiency. A theoretical analysis is presented, demonstrating the superior robustness of the proposed Layerwise Cosine Aggregation compared to original robust aggregation operators. Empirical evaluation across diverse image classification datasets, under varying data distributions and Byzantine attack scenarios, consistently demonstrates the improved performance of Layerwise Cosine Aggregation, achieving up to a 16% increase in model accuracy.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en:&nbsp;<a href=\"https:\/\/arxiv.org\/abs\/2503.21244\">arXiv:2503.21244<\/a>&nbsp;[cs.LG]<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI:&nbsp;<\/strong><a href=\"https:\/\/doi.org\/10.48550\/arXiv.2503.21244\">https:\/\/doi.org\/10.48550\/arXiv.2503.21244<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/arxiv.org\/abs\/2503.21244\">Enlace al art\u00edculo<\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<!--nextpage-->\n\n\n\n<div class=\"wp-block-media-text is-stacked-on-mobile\" style=\"grid-template-columns:40% auto\"><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"685\" height=\"751\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/13748_2025_419_Fig1_HTML.png\" alt=\"\" class=\"wp-image-9673 size-full\" srcset=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/13748_2025_419_Fig1_HTML.png 685w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/13748_2025_419_Fig1_HTML-274x300.png 274w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/13748_2025_419_Fig1_HTML-11x12.png 11w\" sizes=\"(max-width: 685px) 100vw, 685px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo &#8211; Green-EDP: aligning personalization in federated learning and green artificial intelligence throughout the encoder-decoder architecture<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores<\/strong> &#8211; <a href=\"https:\/\/link.springer.com\/article\/10.1007\/s13748-025-00419-3#auth-Jos__Antonio-Ruiz_Mill_n-Aff1\">Jos\u00e9 Antonio Ruiz-Mill\u00e1n<\/a>,&nbsp;<br><a href=\"https:\/\/link.springer.com\/article\/10.1007\/s13748-025-00419-3#auth-Nuria-Rodr_guez_Barroso-Aff1-Aff2\">Nuria Rodr\u00edguez-Barroso<\/a>&nbsp;&amp;&nbsp;<br><a href=\"https:\/\/link.springer.com\/article\/10.1007\/s13748-025-00419-3#auth-M__Victoria-Luz_n-Aff1-Aff3\">M. Victoria Luz\u00f3n<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen<\/strong> &#8211; The rapid advancement of Artificial Intelligence introduces significant challenges related to computational efficiency, data privacy, and distributed data management across diverse environments. Federated Learning (FL) effectively addresses these challenges by enabling decentralized training while simultaneously preserving data privacy, but it often struggles with effective personalization, especially in non-IID (non-Independent and Identically Distributed) data scenarios commonly found in real-world applications. To tackle this issue, we propose Green-EDP, a novel and modular FL architecture that balances global generalization and local adaptation by leveraging an Encoder-Decoder-based architecture. The encoder, hosted on the central server, aggregates shared knowledge from all participating clients, while the decoder, private to each individual client, integrates these global insights with specific local data to enhance personalized model performance. Our method is fully modular and can be flexibly combined with different FL aggregation techniques, optimizers, and various personalization strategies. We evaluate Green-EDP on multiple federated datasets (EMNIST, CelebA, CIFAR-10) and demonstrate that it achieves superior accuracy and significantly faster convergence while maintaining per-round training times comparable to baselines. At the same time, by transmitting only the encoder parameters, Green-EDP reduces the communication cost per round by an order of magnitude, which together with its quicker convergence lowers the overall computational and communication footprint. This dual efficiency aligns Green-EDP with Green AI principles, offering a sustainable and effective solution for personalized federated learning.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en:&nbsp;<a href=\"https:\/\/arxiv.org\/abs\/2503.21244\">arXiv:2503.21244<\/a>&nbsp;[cs.LG]<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI:&nbsp;<\/strong><a href=\"https:\/\/doi.org\/10.1007\/s13748-025-00419-3\">https:\/\/doi.org\/10.1007\/s13748-025-00419-3<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/link.springer.com\/article\/10.1007\/s13748-025-00419-3\">Enlace al art\u00edculo<\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<!--nextpage-->\n\n\n\n<div class=\"wp-block-media-text is-stacked-on-mobile\" style=\"grid-template-columns:40% auto\"><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"552\" height=\"341\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-121212.png\" alt=\"\" class=\"wp-image-9679 size-full\" srcset=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-121212.png 552w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-121212-300x185.png 300w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-121212-18x12.png 18w\" sizes=\"(max-width: 552px) 100vw, 552px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo &#8211; Addressing Data Quality Decompensation in Federated Learning via Dynamic Client Selection<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores<\/strong> &#8211; <a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Fei,+Q\">Qinjun Fei<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Rodr%C3%ADguez-Barroso,+N\">Nuria Rodr\u00edguez-Barroso<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Luz%C3%B3n,+M+V\">Mar\u00eda Victoria Luz\u00f3n<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Zhang,+Z\">Zhongliang Zhang<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Herrera,+F\">Francisco Herrera<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen<\/strong> &#8211; In cross-silo Federated Learning (FL), client selection is critical to ensure high model performance, yet it remains challenging due to data quality decompensation, budget constraints, and incentive compatibility. As training progresses, these factors exacerbate client heterogeneity and degrade global performance. Most existing approaches treat these challenges in isolation, making jointly optimizing multiple factors difficult. To address this, we propose Shapley-Bid Reputation Optimized Federated Learning (SBRO-FL), a unified framework integrating dynamic bidding, reputation modeling, and cost-aware selection. Clients submit bids based on their perceived data quality, and their contributions are evaluated using Shapley values to quantify their marginal impact on the global model. A reputation system, inspired by prospect theory, captures historical performance while penalizing inconsistency. The client selection problem is formulated as a 0-1 integer program that maximizes reputation-weighted utility under budget constraints. Experiments on FashionMNIST, EMNIST, CIFAR-10, and SVHN datasets show that SBRO-FL improves accuracy, convergence speed, and robustness, even in adversarial and low-bid interference scenarios. Our results highlight the importance of balancing data reliability, incentive compatibility, and cost efficiency to enable scalable and trustworthy FL deployments.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en<\/strong> &#8211; <strong><a href=\"https:\/\/arxiv.org\/abs\/2505.21219\">arXiv:2505.21219<\/a>&nbsp;[cs.LG]<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI<\/strong> &#8211; <a href=\"https:\/\/doi.org\/10.48550\/arXiv.2505.21219\">https:\/\/doi.org\/10.48550\/arXiv.2505.21219<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/arxiv.org\/abs\/2505.21219\">Enlace al art\u00edculo<\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<div data-wp-interactive=\"core\/file\" class=\"wp-block-file\"><object data-wp-bind--hidden=\"!state.hasPdfPreview\" hidden class=\"wp-block-file__embed\" data=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Addressing-Data-Quality-Decompensation-in.pdf\" type=\"application\/pdf\" style=\"width:100%;height:600px\" aria-label=\"Embed of Addressing Data Quality Decompensation in.\"><\/object><a id=\"wp-block-file--media-a454c49a-9c2a-44ae-9c6d-312dbd510a61\" href=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Addressing-Data-Quality-Decompensation-in.pdf\">Addressing Data Quality Decompensation in<\/a><a href=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Addressing-Data-Quality-Decompensation-in.pdf\" class=\"wp-block-file__button wp-element-button\" download aria-describedby=\"wp-block-file--media-a454c49a-9c2a-44ae-9c6d-312dbd510a61\">Descarga<\/a><\/div>\n\n\n\n<!--nextpage-->\n\n\n\n<div class=\"wp-block-media-text is-stacked-on-mobile\" style=\"grid-template-columns:40% auto\"><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"587\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/MIR-2024-10-483-1-1-1024x587.jpg\" alt=\"\" class=\"wp-image-9682 size-full\" srcset=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/MIR-2024-10-483-1-1-1024x587.jpg 1024w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/MIR-2024-10-483-1-1-300x172.jpg 300w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/MIR-2024-10-483-1-1-768x441.jpg 768w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/MIR-2024-10-483-1-1-1536x881.jpg 1536w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/MIR-2024-10-483-1-1-2048x1175.jpg 2048w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/MIR-2024-10-483-1-1-18x10.jpg 18w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo &#8211; RAB<sup>2<\/sup>-DEF: Dynamic and Explainable Defense Against Adversarial Attacks in Federated Learning to Fair Poor Clients<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores<\/strong> &#8211; <a href=\"javascript:void(0);\">Nuria Rodr\u00edguez-Barroso<\/a>, <a href=\"javascript:void(0);\">M. Victoria Luz\u00f3n<\/a>, <a href=\"javascript:void(0);\">Francisco Herrera<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen<\/strong> &#8211; When artificial intelligence is becoming popular, the concern and the need for regulation are growing, besides other requirements of the data privacy. In this context, federated learning is proposed as a solution to data privacy concerns derived from different source data scenarios due to its distributed learning. The defense mechanisms proposed in the literature focus only on defending against adversarial attacks and maintaining performance, ignoring other important qualities such as explainability and fairness to poor quality clients, dynamism in terms of attack configuration and generality in terms of being resilient against different kinds of attacks. In this work, we propose RAB<sup>2<\/sup>-DEF, a resilient defense against byzantine and backdoor attacks which is dynamic, explainable and fair to poor clients via local linear explanations. We test the performance of RAB<sup>2<\/sup>-DEF on image datasets and defending against the byzantine and backdoor attacks considering the state-of-the-art defenses, and the result reveals that RAB<sup>2<\/sup>-DEF is a proper defense while also enhancing the other qualities toward trustworthy artificial intelligence<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en<\/strong> &#8211; <em>Machine Intelligence Research<\/em>, 2026, 23(1): 133-146.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI<\/strong> &#8211; <a href=\"https:\/\/dx.doi.org\/10.1007\/s11633-025-1557-1\" target=\"_blank\" rel=\"noreferrer noopener\">10.1007\/s11633-025-1557-1<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/www.mi-research.net\/en\/article\/doi\/10.1007\/s11633-025-1557-1\">Enlace al art\u00edculo<\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<!--nextpage-->\n\n\n\n<div class=\"wp-block-media-text is-stacked-on-mobile\"><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"453\" height=\"350\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-122220.png\" alt=\"\" class=\"wp-image-9684 size-full\" srcset=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-122220.png 453w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-122220-300x232.png 300w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-122220-16x12.png 16w\" sizes=\"(max-width: 453px) 100vw, 453px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo &#8211; An Interpretable Client Decision Tree Aggregation process for Federated Learning<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores <\/strong>&#8211; <a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Argente-Garrido,+A\">Alberto Argente-Garrido<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Zuheros,+C\">Cristina Zuheros<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Luz%C3%B3n,+M+V\">M. Victoria Luz\u00f3n<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Herrera,+F\">Francisco Herrera<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen <\/strong>&#8211; Trustworthy Artificial Intelligence solutions are essential in today&#8217;s data-driven applications, prioritizing principles such as robustness, safety, transparency, explainability, and privacy among others. This has led to the emergence of Federated Learning as a solution for privacy and distributed machine learning. While decision trees, as self-explanatory models, are ideal for collaborative model training across multiple devices in resource-constrained environments such as federated learning environments for injecting interpretability in these models. Decision tree structure makes the aggregation in a federated learning environment not trivial. They require techniques that can merge their decision paths without introducing bias or overfitting while keeping the aggregated decision trees robust and generalizable. In this paper, we propose an Interpretable Client Decision Tree Aggregation process for Federated Learning scenarios that keeps the interpretability and the precision of the base decision trees used for the aggregation. This model is based on aggregating multiple decision paths of the decision trees and can be used on different decision tree types, such as ID3 and CART. We carry out the experiments within four datasets, and the analysis shows that the tree built with the model improves the local models, and outperforms the state-of-the-art.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en <\/strong>&#8211; <strong><a href=\"https:\/\/arxiv.org\/abs\/2404.02510\">arXiv:2404.02510<\/a>&nbsp;[cs.LG]<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI <\/strong>&#8211; <a href=\"https:\/\/doi.org\/10.48550\/arXiv.2404.02510\">ttps:\/\/doi.org\/10.48550\/arXiv.2404.02510<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/arxiv.org\/abs\/2404.02510\">Enlace al art\u00edculo<\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<!--nextpage-->\n\n\n\n<div class=\"wp-block-media-text is-stacked-on-mobile\" style=\"grid-template-columns:40% auto\"><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"497\" height=\"248\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-124720.png\" alt=\"\" class=\"wp-image-9686 size-full\" srcset=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-124720.png 497w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-124720-300x150.png 300w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-124720-18x9.png 18w\" sizes=\"(max-width: 497px) 100vw, 497px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo<\/strong> &#8211; <strong>Improving Krum\u2019s Byzantine Resilience in Federated Learning via layerwise aggregation<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores<\/strong> &#8211; <a href=\"https:\/\/ieeexplore.ieee.org\/author\/269565890064084\">Mario Garc\u00eda-M\u00e1rquez<\/a>,&nbsp;<a href=\"https:\/\/ieeexplore.ieee.org\/author\/37087027006\">Nuria Rodr\u00edguez-Barroso<\/a>,&nbsp;<a href=\"https:\/\/ieeexplore.ieee.org\/author\/38196461900\">M. Victoria Luz\u00f3n<\/a>,&nbsp;<a href=\"https:\/\/ieeexplore.ieee.org\/author\/37270827700\">Francisco Herrera<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen<\/strong> &#8211; With the rising usage of artificial intelligence systems, social concerns around them and the need for regulation is also increasing, including requirements for data privacy. Federated Learning addresses data privacy concerns in distributed machine learning by training models collaboratively without centralizing data. However, Federated Learning is susceptible to Byzantine attacks, where malicious nodes submit corrupted updates. Krum, a robust aggregation algorithm, has been widely adopted as a defense mechanism. However, recent studies have shown that Krum\u2019s performance degrades significantly in high-dimensional settings. This work proposes Layerwise Krum, a novel aggregation method that enhances Krum\u2019s robustness in high-dimensional spaces while maintaining computational efficiency. We provide theoretical analysis of the improved robustness of Layerwise Krum compared to standard Krum. Furthermore, we empirically evaluate Layerwise Krum on various image classification datasets under diverse data distributions and Byzantine attack scenarios, consistently demonstrating superior performance compared to the original Krum operator.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en<\/strong> &#8211; <a href=\"https:\/\/ieeexplore.ieee.org\/xpl\/conhome\/11227166\/proceeding\">2025 International Joint Conference on Neural Networks (IJCNN)<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI <\/strong>&#8211; <strong>&nbsp;<\/strong><a href=\"https:\/\/doi.org\/10.1109\/IJCNN64981.2025.11228388\" target=\"_blank\" rel=\"noreferrer noopener\">10.1109\/IJCNN64981.2025.11228388<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/ieeexplore.ieee.org\/document\/11228388\">Enlace al art\u00edculo<\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<!--nextpage-->\n\n\n\n<div class=\"wp-block-media-text is-stacked-on-mobile\"><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"582\" height=\"223\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-125235.png\" alt=\"\" class=\"wp-image-9688 size-full\" srcset=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-125235.png 582w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-125235-300x115.png 300w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-125235-18x7.png 18w\" sizes=\"(max-width: 582px) 100vw, 582px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo &#8211; Security Threats to Explainable Classifiers in Federated Learning<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores<\/strong> &#8211; <a href=\"https:\/\/ieeexplore.ieee.org\/author\/37090078645\">Mattia Daole<\/a>, <a href=\"https:\/\/ieeexplore.ieee.org\/author\/37642939200\">Pietro Ducange<\/a>,&nbsp;<a href=\"https:\/\/ieeexplore.ieee.org\/author\/37270827700\">Francisco Herrera<\/a>,&nbsp;<a href=\"https:\/\/ieeexplore.ieee.org\/author\/37284359200\">Francesco Marcelloni<\/a>,&nbsp;<a href=\"https:\/\/ieeexplore.ieee.org\/author\/37087041652\">Alessandro Renda<\/a>,&nbsp;<a href=\"https:\/\/ieeexplore.ieee.org\/author\/37087027006\">Nuria Rodr\u00edguez-Barroso<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen <\/strong>&#8211; The decentralized nature of federated learning (FL) poses critical challenges related to security: Clients participating in the process may not necessarily be trustworthy and could engage in adversarial attacks, potentially undermining the integrity and reliability of the global machine learning model. Security concerns have been extensively investigated in traditional FL, where collaboratively learned models are typically deep neural networks. However, this class of models does not meet the requirement of explainability, which is considered essential for the trustworthiness of AI systems. In this work, we present an analysis on security threats to FL of explainable models, namely fuzzy rule-based classifiers (FRBCs). We outline the types of attacks a malicious client may implement, and assess, through a preliminary experimental analysis, the impact they have on FL of FRBCs in terms of global model performance. We also compare these findings with the effects of the same or similar well-established attacks in traditional FL of neural network models. Finally, we provide insights to improve the security of FRBCs learned in a federated fashion.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en<\/strong> &#8211; <a href=\"https:\/\/ieeexplore.ieee.org\/xpl\/conhome\/11227166\/proceeding\">2025 International Joint Conference on Neural Networks (IJCNN)<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI <\/strong>&#8211; <a href=\"https:\/\/doi.org\/10.1109\/IJCNN64981.2025.11227961\" target=\"_blank\" rel=\"noreferrer noopener\">10.1109\/IJCNN64981.2025.11227961<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/ieeexplore.ieee.org\/document\/11227961\">Enlace al art\u00edculo<\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<!--nextpage-->\n\n\n\n<div class=\"wp-block-media-text is-stacked-on-mobile\" style=\"grid-template-columns:40% auto\"><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"343\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-131439-1024x343.png\" alt=\"\" class=\"wp-image-9693 size-full\" srcset=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-131439-1024x343.png 1024w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-131439-300x101.png 300w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-131439-768x257.png 768w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-131439-18x6.png 18w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-131439.png 1262w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo &#8211; Establishing the&nbsp;Foundation for&nbsp;Out-of-Distribution Detection in&nbsp;Monument Classification Through Nested Dichotomies<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores<\/strong> &#8211; <a href=\"https:\/\/link.springer.com\/chapter\/10.1007\/978-3-031-74186-9_14#auth-Ignacio-Antequera_S_nchez\">Ignacio Antequera-S\u00e1nchez<\/a>,&nbsp;<a href=\"https:\/\/link.springer.com\/chapter\/10.1007\/978-3-031-74186-9_14#auth-Juan_Luis-Su_rez_D_az\">Juan Luis Su\u00e1rez-D\u00edaz<\/a>,&nbsp;<a href=\"https:\/\/link.springer.com\/chapter\/10.1007\/978-3-031-74186-9_14#auth-Rosana-Montes\">Rosana Montes<\/a>, <a href=\"https:\/\/link.springer.com\/chapter\/10.1007\/978-3-031-74186-9_14#auth-Francisco-Herrera\">Francisco Herrera<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen <\/strong>&#8211; This paper introduces a hierarchical approach utilizing nested dichotomies to enhance the MonuMAI framework designed&nbsp;for architectural image classification. The study focuses on developing a foundational layer dedicated to distinguishing between building and non-building images, effectively reducing complexity&nbsp;by filtering out irrelevant data early in the classification process. Through empirical investigation utilizing a fine-tuned EfficientNet model, the study demonstrates substantial progress in handling out-of-distribution scenarios in monument detection. The&nbsp;model effectively filters out irrelevant images while accurately retaining most monument images, establishing a robust foundation&nbsp;for subsequent layers aimed at improving out-of-distribution detection, and recognizing new architectural styles post-deployment. This&nbsp;study specifically emphasizes the initial layer of our innovative system, setting the stage for future expansion and development of subsequent layers. This research marks a significant stride in mitigating&nbsp;OOD challenges within architectural image classification, highlighting the potential of hierarchical methodologies to propel MonuMAI&nbsp;and analogous systems towards more precise and adaptable AI solutions.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en <\/strong>&#8211; H. Quinti\u00b4an et al. (Eds.): HAIS 2024, LNAI 14858, pp. 165\u2013176, 2025<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI <\/strong>&#8211; <a href=\"https:\/\/doi.org\/10.1007\/978-3-031-74186-9_14\">https:\/\/doi.org\/10.1007\/978-3-031-74186-9_14<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/link.springer.com\/chapter\/10.1007\/978-3-031-74186-9_14\">Enlace al art\u00edculo<\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<!--nextpage-->\n\n\n\n<div class=\"wp-block-media-text is-stacked-on-mobile\" style=\"grid-template-columns:40% auto\"><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"568\" height=\"240\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/1-s2.0-S0952197625036255-gr1.jpg\" alt=\"\" class=\"wp-image-9696 size-full\" srcset=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/1-s2.0-S0952197625036255-gr1.jpg 568w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/1-s2.0-S0952197625036255-gr1-300x127.jpg 300w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/1-s2.0-S0952197625036255-gr1-18x8.jpg 18w\" sizes=\"(max-width: 568px) 100vw, 568px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo &#8211; A task-oriented few-shot spectroscopy benchmark: Empirical evaluation of meta-learning approaches<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores <\/strong>&#8211; Iv\u00e1n&nbsp;Garz\u00f3n,&nbsp;Guillermo&nbsp;Gomez-Trenado,&nbsp;Sergio&nbsp;Damas<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen<\/strong> &#8211; Near- and mid-infrared (NIR\/MIR) spectroscopy, crucial for rapid, noninvasive chemical analysis in industrial applications, is increasingly leveraged by machine learning techniques to extract meaningful information from complex spectral data. Nevertheless, the constraints imposed by limited sample sizes and the occurrence of domain shifts present substantial obstacles to the effectiveness of both deep learning and transfer learning methodologies. In this work, we introduce a unified, task-oriented benchmark that consolidates nine open-access spectral datasets, providing a consistent preprocessing pipeline and clearly defined few-shot tasks. The benchmark is segmented into multiple tests exhibiting varying degrees of heterogeneity, thereby facilitating the evaluation of diverse methodologies across a range of contexts characterized by varying levels of complexity. To the best of our knowledge, meta-learning has not previously been applied in the context of spectrometric analysis. In this study, we confirm that it offers superior performance to traditional transfer learning approaches. A systematic comparison of transfer learning and meta-learning techniques reveals that meta-learning consistently outperforms transfer learning\u2014achieving lower errors in both heterogeneous and homogeneous tasks, consistent with our latent-space analysis showing a more structured cross-task embedding geometry that helps explain these gains. Furthermore, we empirically demonstrate that this benchmark can serve as a starting point for pre-training models on new industrial problems, accelerating adaptation and improving final performance using meta-learning. We complement our analysis with established, model-agnostic interpretability tools, enhancing transparency and industrial relevance in regulated domains.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en<\/strong> &#8211; <a href=\"https:\/\/www.sciencedirect.com\/journal\/engineering-applications-of-artificial-intelligence\">Engineering Applications of Artificial Intelligence<\/a> <a href=\"https:\/\/www.sciencedirect.com\/journal\/engineering-applications-of-artificial-intelligence\/vol\/167\/part\/P3\">Volume 167, Part 3<\/a>,&nbsp;1 March 2026, 113593<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI <\/strong> &#8211; h<a href=\"https:\/\/doi.org\/10.1016\/j.engappai.2025.113593\">ttps:\/\/doi.org\/10.1016\/j.engappai.2025.113593<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/www.sciencedirect.com\/science\/article\/pii\/S0952197625036255#fig1\">Enlace al art\u00edculo<\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<!--nextpage-->\n\n\n\n<div class=\"wp-block-media-text is-stacked-on-mobile\" style=\"grid-template-columns:40% auto\"><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"446\" height=\"217\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-133710.png\" alt=\"\" class=\"wp-image-9698 size-full\" srcset=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-133710.png 446w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-133710-300x146.png 300w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-133710-18x9.png 18w\" sizes=\"(max-width: 446px) 100vw, 446px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo &#8211; DCV-ROOD Evaluation Framework: Dual Cross-Validation for Robust Out-of-Distribution Detection<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores<\/strong> &#8211; <a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Urrea-Casta%C3%B1o,+A\">Arantxa Urrea-Casta\u00f1o<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Segura-Kunsagi,+N\">Nicol\u00e1s Segura-Kunsagi<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Su%C3%A1rez-D%C3%ADaz,+J+L\">Juan Luis Su\u00e1rez-D\u00edaz<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Montes,+R\">Rosana Montes<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Herrera,+F\">Francisco Herrera<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen<\/strong> &#8211; Out-of-distribution (OOD) detection plays a key role in enhancing the robustness of artificial intelligence systems by identifying inputs that differ significantly from the training distribution, thereby preventing unreliable predictions and enabling appropriate fallback mechanisms. Developing reliable OOD detection methods is a significant challenge, and rigorous evaluation of these techniques is essential for ensuring their effectiveness, as it allows researchers to assess their performance under diverse conditions and to identify potential limitations or failure modes. Cross-validation (CV) has proven to be a highly effective tool for providing a reasonable estimate of the performance of a learning algorithm. Although OOD scenarios exhibit particular characteristics, an appropriate adaptation of CV can lead to a suitable evaluation framework for this setting. This work proposes a dual CV framework for robust evaluation of OOD detection models, aimed at improving the reliability of their assessment. The proposed evaluation framework aims to effectively integrate in-distribution (ID) and OOD data while accounting for their differing characteristics. To achieve this, ID data are partitioned using a conventional approach, whereas OOD data are divided by grouping samples based on their classes. Furthermore, we analyze the context of data with class hierarchy to propose a data splitting that considers the entire class hierarchy to obtain fair ID-OOD partitions to apply the proposed evaluation framework. This framework is called Dual Cross-Validation for Robust Out-of-Distribution Detection (DCV-ROOD). To test the validity of the evaluation framework, we selected a set of state-of-the-art OOD detection methods, both with and without outlier exposure. The results show that the method achieves very fast convergence to the true performance.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en<\/strong> &#8211; <strong><a href=\"https:\/\/arxiv.org\/abs\/2509.05778\">arXiv:2509.05778<\/a>&nbsp;[cs.LG]<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI <\/strong> &#8211; <a href=\"https:\/\/doi.org\/10.48550\/arXiv.2509.05778\">https:\/\/doi.org\/10.48550\/arXiv.2509.05778<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/arxiv.org\/abs\/2509.05778\">Enlace al art\u00edculo<\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<div data-wp-interactive=\"core\/file\" class=\"wp-block-file aligncenter\"><object data-wp-bind--hidden=\"!state.hasPdfPreview\" hidden class=\"wp-block-file__embed\" data=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/DCV-ROOD-Evaluation-Framework-Dual-Cross-Validation-for-Robust-Out-of-Distribution-Detection.pdf\" type=\"application\/pdf\" style=\"width:100%;height:600px\" aria-label=\"Embed of DCV ROOD Evaluation Framework Dual Cross Validation for Robust Out of Distribution Detection.\"><\/object><a id=\"wp-block-file--media-342de612-cdb2-44ee-a4de-1733f07e83b1\" href=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/DCV-ROOD-Evaluation-Framework-Dual-Cross-Validation-for-Robust-Out-of-Distribution-Detection.pdf\">DCV ROOD Evaluation Framework Dual Cross Validation for Robust Out of Distribution Detection<\/a><a href=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/DCV-ROOD-Evaluation-Framework-Dual-Cross-Validation-for-Robust-Out-of-Distribution-Detection.pdf\" class=\"wp-block-file__button wp-element-button\" download aria-describedby=\"wp-block-file--media-342de612-cdb2-44ee-a4de-1733f07e83b1\">Descarga<\/a><\/div>\n\n\n\n<!--nextpage-->\n\n\n\n<div class=\"wp-block-media-text is-stacked-on-mobile\" style=\"grid-template-columns:40% auto\"><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"685\" height=\"416\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/631509_1_En_15_Fig1_HTML.png\" alt=\"\" class=\"wp-image-9699 size-full\" srcset=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/631509_1_En_15_Fig1_HTML.png 685w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/631509_1_En_15_Fig1_HTML-300x182.png 300w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/631509_1_En_15_Fig1_HTML-18x12.png 18w\" sizes=\"(max-width: 685px) 100vw, 685px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo &#8211; A Preliminary Study on&nbsp;Preprocessing the&nbsp;Semantic Space in&nbsp;Zero-Shot Learning<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores<\/strong> &#8211; Juan Jos\u00e9 Herrera Aranda, Francisco Herrera, Isaac Triguero<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen<\/strong> &#8211; Realising a general purpose artificial intelligence is&nbsp;an ambitious goal. To do this, new techniques, especially in&nbsp;an open-world context where new tasks may appear, are to be developed to deal with the dynamism and diversity of these systems. A relevant example of this kind is Zero-Shot Learning, where training and&nbsp;test sets have disjoint label sets, so that, no examples of the classes expected in the test (unseen classes) are available during&nbsp;the training phase, and the training (seen) classes are not involved&nbsp;in the test phase. Recent studies aim to generalise from seen to unseen classes by building a variety of models that exploit semantic information in addition to the actual samples (e.g. images). However, the adequacy of the input semantic attributes has not&nbsp;been explored. This initial work studies the influence of preprocessing the semantic space using embedded feature selection within&nbsp;a cross-validation scheme to generalise to unseen classes.&nbsp;The experiments on three well-known zero-shot learning benchmarks&nbsp;show that selecting a subset of the semantic attributes may be beneficial to learning more general methods. <\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en<\/strong> &#8211; H. Quinti\u00b4an et al. (Eds.): HAIS 2024, LNAI 14858, pp. 177-189, 2025<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI<\/strong> &#8211; <a href=\"https:\/\/doi.org\/10.1007\/978-3-031-74186-9_15\">https:\/\/doi.org\/10.1007\/978-3-031-74186-9_15<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/link.springer.com\/chapter\/10.1007\/978-3-031-74186-9_15\">Enlace al art\u00edculo<\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<!--nextpage-->\n\n\n\n<div class=\"wp-block-media-text is-stacked-on-mobile\" style=\"grid-template-columns:40% auto\"><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"210\" height=\"133\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-134630.png\" alt=\"\" class=\"wp-image-9702 size-full\" srcset=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-134630.png 210w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-134630-18x12.png 18w\" sizes=\"(max-width: 210px) 100vw, 210px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo &#8211; A First Approach to Refine Semantic Spaces in Zero-Shot Learning with a Genetic Algorithm<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores <\/strong>&#8211; Juan Jos\u00e9 Herrera Aranda, Francisco Herrera, Isaac Triguero<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen<\/strong> &#8211; Evolutionary computation has been successfully applied to tackle a wide variety of machine learning problems due to its generalisation and adaptability capabilities. Recently, it has shown great potential to enhance General Purpose Artificial<br>Intelligence Systems particularly, those that work in open-world scenarios which require dynamic adaptation abilities. Zero-Shot Learning (ZSL) is an emerging paradigm within the open-world context that allows us to perform predictive tasks, such as the classification of unknown elements (i.e. unknown classes) for which a model has not been specifically trained. To do this, ZSL<br>uses auxiliary information known as semantic space, typically in the form of attributes that define each class. This plays a crucial role in associating prior knowledge with unknown situations. However, the treatment of the semantic space has remained an underexplored area, as selecting the most relevant semantic attributes that generalise to unknown classes is a challenging problem. In this preliminary work, we propose a tailored genetic<br>algorithm to perform feature selection of the semantic space, removing irrelevant features that negatively affect the generalisation capabilities of a well-known ZSL approach. The results on four commonly used ZSL image classification problems show that refining the semantic space may consistently boost the accuracy across all datasets.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en<\/strong> &#8211; <em>2025 IEEE Congress on Evolutionary Computation (CEC)<\/em>, Hangzhou, China, 2025, pp. 1-4,<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI <\/strong>&#8211; <a href=\"https:\/\/ieeexplore.ieee.org\/document\/11042972\">10.1109\/CEC65147.2025.11042972<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/ieeexplore.ieee.org\/document\/11042972\">Enlace al art\u00edculo<\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<!--nextpage-->\n\n\n\n<div class=\"wp-block-media-text is-stacked-on-mobile\" style=\"grid-template-columns:40% auto\"><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"472\" height=\"167\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-135431.png\" alt=\"\" class=\"wp-image-9704 size-full\" srcset=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-135431.png 472w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-135431-300x106.png 300w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-135431-18x6.png 18w\" sizes=\"(max-width: 472px) 100vw, 472px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo &#8211; Semantic-Inductive Attribute Selection for Zero-Shot Learning<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores<\/strong> &#8211; <a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Herrera-Aranda,+J+J\">Juan Jose Herrera-Aranda<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Gomez-Trenado,+G\">Guillermo Gomez-Trenado<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Herrera,+F\">Francisco Herrera<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Triguero,+I\">Isaac Triguero<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen<\/strong> &#8211; Zero-Shot Learning is an important paradigm within General-Purpose Artificial Intelligence Systems, particularly in those that operate in open-world scenarios where systems must adapt to new tasks dynamically. Semantic spaces play a pivotal role as they bridge seen and unseen classes, but whether human-annotated or generated by a machine learning model, they often contain noisy, redundant, or irrelevant attributes that hinder performance. To address this, we introduce a partitioning scheme that simulates unseen conditions in an inductive setting (which is the most challenging), allowing attribute relevance to be assessed without access to semantic information from unseen classes. Within this framework, we study two complementary feature-selection strategies and assess their generalisation. The first adapts embedded feature selection to the particular demands of ZSL, turning model-driven rankings into meaningful semantic pruning; the second leverages evolutionary computation to directly explore the space of attribute subsets more broadly. Experiments on five benchmark datasets (AWA2, CUB, SUN, aPY, FLO) show that both methods consistently improve accuracy on unseen classes by reducing redundancy, but in complementary ways: RFS is efficient and competitive though dependent on critical hyperparameters, whereas GA is more costly yet explores the search space more broadly and avoids such dependence. These results confirm that semantic spaces are inherently redundant and highlight the proposed partitioning scheme as an effective tool to refine them under inductive conditions.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en<\/strong> &#8211; <strong><a href=\"https:\/\/arxiv.org\/abs\/2510.03260\">arXiv:2510.03260<\/a>&nbsp;[cs.LG]<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI <\/strong>&#8211; <a href=\"https:\/\/doi.org\/10.48550\/arXiv.2510.03260\">https:\/\/doi.org\/10.48550\/arXiv.2510.03260<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/arxiv.org\/abs\/2510.03260\">Enlace al art\u00edculo<\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<div data-wp-interactive=\"core\/file\" class=\"wp-block-file aligncenter\"><object data-wp-bind--hidden=\"!state.hasPdfPreview\" hidden class=\"wp-block-file__embed\" data=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Semantic-Inductive-Attribute-Selection-for-Zero-Shot-Learning.pdf\" type=\"application\/pdf\" style=\"width:100%;height:600px\" aria-label=\"Embed of Semantic-Inductive Attribute Selection for Zero-Shot Learning.\"><\/object><a id=\"wp-block-file--media-34904331-feeb-489c-9fef-1a7adef72200\" href=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Semantic-Inductive-Attribute-Selection-for-Zero-Shot-Learning.pdf\">Semantic-Inductive Attribute Selection for Zero-Shot Learning<\/a><a href=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Semantic-Inductive-Attribute-Selection-for-Zero-Shot-Learning.pdf\" class=\"wp-block-file__button wp-element-button\" download aria-describedby=\"wp-block-file--media-34904331-feeb-489c-9fef-1a7adef72200\">Descarga<\/a><\/div>\n\n\n\n<!--nextpage-->\n\n\n\n<div class=\"wp-block-media-text is-stacked-on-mobile\" style=\"grid-template-columns:40% auto\"><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"685\" height=\"321\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/631509_1_En_10_Fig2_HTML.png\" alt=\"\" class=\"wp-image-9706 size-full\" srcset=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/631509_1_En_10_Fig2_HTML.png 685w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/631509_1_En_10_Fig2_HTML-300x141.png 300w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/631509_1_En_10_Fig2_HTML-18x8.png 18w\" sizes=\"(max-width: 685px) 100vw, 685px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo &#8211; An In-depth Analysis of&nbsp;Jailbreaking Through Domain Characterization of&nbsp;LLM Training Sets<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores<\/strong> &#8211; Carlos Pel\u00e1ez-Gonz\u00e1lez, Andr\u00e9s Herrera-Poyatos, Francisco Herrera-Triguero<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen<\/strong> &#8211; Research on large language models (LLMs) is a prominent field in open-world machine learning. Despite their significant capabilities in natural language processing, LLMs face several challenges that must be overcome, namely, consistency, hallucinations and jailbreaking. In this work we focus on&nbsp;the latter. Jailbreak attacks involve crafting prompts designed&nbsp;to bypass the model alignment safeguards of LLMs, leading to harmful outputs that compromise the safety of the LLM model. Our&nbsp;main contribution is a fundamental vision of jailbreaking in terms of&nbsp;the different domains of language that arise when training and aligning LLMs. This theoretical perspective showcases the limitations&nbsp;of current methods and allows us to classify jailbreak attacks in&nbsp;terms of which deficiency of the model they exploit. This contrasts&nbsp;to current classifications that are based on how the prompt&nbsp;is constructed, such as prompt templating. We conclude that a deeper understanding of the behavior of LLMs is essential to prevent jailbreak attacks.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en<\/strong> &#8211; H. Quinti\u00b4an et al. (Eds.): HAIS 2024, LNAI 14858, pp. 116-127, 2025<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI <\/strong>&#8211; <a href=\"https:\/\/doi.org\/10.1007\/978-3-031-74186-9_10\">https:\/\/doi.org\/10.1007\/978-3-031-74186-9_10<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/link.springer.com\/chapter\/10.1007\/978-3-031-74186-9_10\">Enlace al art\u00edculo<\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<!--nextpage-->\n\n\n\n<div class=\"wp-block-media-text is-stacked-on-mobile\" style=\"grid-template-columns:40% auto\"><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"599\" height=\"255\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-140206.png\" alt=\"\" class=\"wp-image-9710 size-full\" srcset=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-140206.png 599w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-140206-300x128.png 300w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-140206-18x8.png 18w\" sizes=\"(max-width: 599px) 100vw, 599px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo &#8211; A Domain-Based Taxonomy of Jailbreak Vulnerabilities in Large Language Models<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores <\/strong>&#8211; <a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Pel%C3%A1ez-Gonz%C3%A1lez,+C\">Carlos Pel\u00e1ez-Gonz\u00e1lez<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Herrera-Poyatos,+A\">Andr\u00e9s Herrera-Poyatos<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Zuheros,+C\">Cristina Zuheros<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Herrera-Poyatos,+D\">David Herrera-Poyatos<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Tejedor,+V\">Virilo Tejedor<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Herrera,+F\">Francisco Herrera<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen <\/strong>&#8211; The study of large language models (LLMs) is a key area in open-world machine learning. Although LLMs demonstrate remarkable natural language processing capabilities, they also face several challenges, including consistency issues, hallucinations, and jailbreak vulnerabilities. Jailbreaking refers to the crafting of prompts that bypass alignment safeguards, leading to unsafe outputs that compromise the integrity of LLMs. This work specifically focuses on the challenge of jailbreak vulnerabilities and introduces a novel taxonomy of jailbreak attacks grounded in the training domains of LLMs. It characterizes alignment failures through generalization, objectives, and robustness gaps. Our primary contribution is a perspective on jailbreak, framed through the different linguistic domains that emerge during LLM training and alignment. This viewpoint highlights the limitations of existing approaches and enables us to classify jailbreak attacks on the basis of the underlying model deficiencies they exploit. Unlike conventional classifications that categorize attacks based on prompt construction methods (e.g., prompt templating), our approach provides a deeper understanding of LLM behavior. We introduce a taxonomy with four categories &#8212; mismatched generalization, competing objectives, adversarial robustness, and mixed attacks &#8212; offering insights into the fundamental nature of jailbreak vulnerabilities. Finally, we present key lessons derived from this taxonomic study.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en <\/strong>&#8211; <strong><a href=\"https:\/\/arxiv.org\/abs\/2504.04976\">arXiv:2504.04976<\/a>&nbsp;[cs.CL]<\/strong> <\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI<\/strong> &#8211; <a href=\"https:\/\/doi.org\/10.48550\/arXiv.2504.04976\">https:\/\/doi.org\/10.48550\/arXiv.2504.04976<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/arxiv.org\/abs\/2504.04976\">Enlace al art\u00edculo <\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<div data-wp-interactive=\"core\/file\" class=\"wp-block-file aligncenter\"><object data-wp-bind--hidden=\"!state.hasPdfPreview\" hidden class=\"wp-block-file__embed\" data=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/A-Domain-Based-Taxonomy-of-Jailbreak-Vulnerabilities-in-Large-Language-Models.pdf\" type=\"application\/pdf\" style=\"width:100%;height:600px\" aria-label=\"Embed of A Domain-Based Taxonomy of Jailbreak Vulnerabilities in Large Language Models.\"><\/object><a id=\"wp-block-file--media-f442a080-cf10-4178-a3e8-593a0eedd671\" href=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/A-Domain-Based-Taxonomy-of-Jailbreak-Vulnerabilities-in-Large-Language-Models.pdf\">A Domain-Based Taxonomy of Jailbreak Vulnerabilities in Large Language Models<\/a><a href=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/A-Domain-Based-Taxonomy-of-Jailbreak-Vulnerabilities-in-Large-Language-Models.pdf\" class=\"wp-block-file__button wp-element-button\" download aria-describedby=\"wp-block-file--media-f442a080-cf10-4178-a3e8-593a0eedd671\">Descarga<\/a><\/div>\n\n\n\n<!--nextpage-->\n\n\n\n<div class=\"wp-block-media-text is-stacked-on-mobile\" style=\"grid-template-columns:40% auto\"><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"632\" height=\"357\" src=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-141002.png\" alt=\"\" class=\"wp-image-9713 size-full\" srcset=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-141002.png 632w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-141002-300x169.png 300w, https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/Captura-de-pantalla-2026-02-19-141002-18x10.png 18w\" sizes=\"(max-width: 632px) 100vw, 632px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>T\u00edtulo &#8211; SeNeDiF-OOD: Semantic Nested Dichotomy Fusion for Out-of-Distribution Detection Methodology in Open-World Classification. A Case Study on Monument Style Classification<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Autores<\/strong> &#8211; <a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Antequera-S%C3%A1nchez,+I\">Ignacio Antequera-S\u00e1nchez<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Su%C3%A1rez-D%C3%ADaz,+J+L\">Juan Luis Su\u00e1rez-D\u00edaz<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Montes,+R\">Rosana Montes<\/a>,&nbsp;<a href=\"https:\/\/arxiv.org\/search\/cs?searchtype=author&amp;query=Herrera,+F\">Francisco Herrera<\/a> <\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Resumen<\/strong> &#8211; Out-of-distribution (OOD) detection is a fundamental requirement for the reliable deployment of artificial intelligence applications in open-world environments. However, addressing the heterogeneous nature of OOD data, ranging from low-level corruption to semantic shifts, remains a complex challenge that single-stage detectors often fail to resolve. To address this issue, we propose SeNeDiF-OOD, a novel methodology based on Semantic Nested Dichotomy Fusion. This framework decomposes the detection task into a hierarchical structure of binary fusion nodes, where each layer is designed to integrate decision boundaries aligned with specific levels of semantic abstraction. To validate the proposed framework, we present a comprehensive case study using MonuMAI, a real-world architectural style recognition system exposed to an open environment. This application faces a diverse range of inputs, including non-monument images, unknown architectural styles, and adversarial attacks, making it an ideal testbed for our proposal. Through extensive experimental evaluation in this domain, results demonstrate that our hierarchical fusion methodology significantly outperforms traditional baselines, effectively filtering these diverse OOD categories while preserving in-distribution performance.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Publicado en<\/strong> &#8211; <strong><a href=\"https:\/\/arxiv.org\/abs\/2601.18739\">arXiv:2601.18739<\/a>&nbsp;[cs.CV]<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DOI<\/strong> &#8211; <a href=\"https:\/\/doi.org\/10.48550\/arXiv.2601.18739\">https:\/\/doi.org\/10.48550\/arXiv.2601.18739<\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/www.arxiv.org\/abs\/2601.18739\">Enlace al art\u00edculo<\/a><\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<div data-wp-interactive=\"core\/file\" class=\"wp-block-file aligncenter\"><object data-wp-bind--hidden=\"!state.hasPdfPreview\" hidden class=\"wp-block-file__embed\" data=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/SeNeDiF-OOD-Semantic-Nested-Dichotomy-Fusion-for-Out-of-Distribution-Detection-Methodology-in-Open-World-Classification.-A-Case-Study-on-Monument-Style-Classification.pdf\" type=\"application\/pdf\" style=\"width:100%;height:600px\" aria-label=\"Embed of SeNeDiF OOD  Semantic Nested Dichotomy Fusion for Out of Distribution Detection Methodology in Open World Classification. A Case Study on Monument Style Classification.\"><\/object><a id=\"wp-block-file--media-c67dbb0f-402b-4e5e-9a62-596401549b1f\" href=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/SeNeDiF-OOD-Semantic-Nested-Dichotomy-Fusion-for-Out-of-Distribution-Detection-Methodology-in-Open-World-Classification.-A-Case-Study-on-Monument-Style-Classification.pdf\">SeNeDiF OOD  Semantic Nested Dichotomy Fusion for Out of Distribution Detection Methodology in Open World Classification. A Case Study on Monument Style Classification<\/a><a href=\"https:\/\/incibe.dasci.es\/wp-content\/uploads\/2026\/02\/SeNeDiF-OOD-Semantic-Nested-Dichotomy-Fusion-for-Out-of-Distribution-Detection-Methodology-in-Open-World-Classification.-A-Case-Study-on-Monument-Style-Classification.pdf\" class=\"wp-block-file__button wp-element-button\" download aria-describedby=\"wp-block-file--media-c67dbb0f-402b-4e5e-9a62-596401549b1f\">Descarga<\/a><\/div>\n","protected":false},"excerpt":{"rendered":"","protected":false},"author":1,"featured_media":0,"parent":89,"menu_order":20,"comment_status":"closed","ping_status":"closed","template":"","meta":{"footnotes":""},"class_list":["post-105","page","type-page","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/incibe.dasci.es\/en\/wp-json\/wp\/v2\/pages\/105","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/incibe.dasci.es\/en\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/incibe.dasci.es\/en\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/incibe.dasci.es\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/incibe.dasci.es\/en\/wp-json\/wp\/v2\/comments?post=105"}],"version-history":[{"count":12,"href":"https:\/\/incibe.dasci.es\/en\/wp-json\/wp\/v2\/pages\/105\/revisions"}],"predecessor-version":[{"id":9773,"href":"https:\/\/incibe.dasci.es\/en\/wp-json\/wp\/v2\/pages\/105\/revisions\/9773"}],"up":[{"embeddable":true,"href":"https:\/\/incibe.dasci.es\/en\/wp-json\/wp\/v2\/pages\/89"}],"wp:attachment":[{"href":"https:\/\/incibe.dasci.es\/en\/wp-json\/wp\/v2\/media?parent=105"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}