https://hdl.handle.net/10259/3847 2026-06-24T00:18:52Z 2026-06-24T00:18:52Z Martin Reizabal, Sergio Caballero Quiroga, Adrian Gil Arroyo, Beatriz Basurto Hornillos, Nuño Ruiz González, Rubén https://hdl.handle.net/10259/11860 2026-06-19T05:15:47Z 2026-03-01T00:00:00Z

Transformer-based classification of IoT network traffic with flow-to-window aggregation Martin Reizabal, Sergio; Caballero Quiroga, Adrian; Gil Arroyo, Beatriz; Basurto Hornillos, Nuño; Ruiz González, Rubén The explosive growth of the IoT has led to an increasingly complex and heterogeneous network traffic, posing major challenges for intrusion detection. Most existing machine learning and deep learning approaches model network traffic at the level of individual flows, which limits their ability to capture contextual relationships among concurrent communications. This paper introduces a Transformer-based framework for IoT intrusion detection that aggregates network flows into fixed-duration windows and treats each flow as a token within the input sequence. The self-attention mechanism captures contextual relationships among concurrent flows, enabling effective modeling of temporal dependencies without recurrence. Experiments conducted on the CICIoT2023 dataset show that the proposed model achieves a weighted F1-score of 97.9% and a macro ROC–AUC of 99.6% under temporally blocked cross-validation, while maintaining high computational efficiency. These results demonstrate that flow-to-window aggregation combined with self-attention provides a robust and scalable foundation for IoT network security, suitable for deployment in edge and smart-home environments.

2026-03-01T00:00:00Z Sánchez Alonso, Ramón Rad Moradillo, Juan Carlos Cambra Baseca, Carlos Barros García, Rocío Herrero Cosío, Álvaro https://hdl.handle.net/10259/11732 2026-05-27T00:05:32Z 2025-12-01T00:00:00Z

Detecting copper-based fungicides in vineyards by means of hyperspectral imagery Sánchez Alonso, Ramón; Rad Moradillo, Juan Carlos; Cambra Baseca, Carlos; Barros García, Rocío; Herrero Cosío, Álvaro Fungal diseases affecting vineyards are commonly controlled using copper-based fungicides. Inaccurate application of these products usually leads to accumulations of copper in the soil. The use of spectral images in vineyards is a tool that can help in the correct application of fungicides to improve their efficiency and effectiveness. To do that, a solution is required to identify the copper deposited on the vine leaf. To bridge this gap, the present work compares images obtained with a hyperspectral camera (Pika L, Resonon) of vineyard leaves (Vitis vinifera L.) cv. Tempranillo treated with two copper-based products, Cuprantol duo (Syngenta, CH) and Cuprocol (Syngenta, CH). Treated leaves with both products and the corresponding blanks made with distilled water were compared. Most of the differences between treatments and products are found in the near-infrared region (700–740 nm), the green region (550 nm) and the region of (620–640 nm). Maximal spectral variations appeared in the range of 711.16–758.27 nm for wet status products, which allowed to differentiate between the areas treated with copper-based products from the blanks without product. We can conclude that using hyperspectral imagery is possible to detect leave areas treated with copper-based fungicides immediately (wet treatment) after application

2025-12-01T00:00:00Z Sánchez Alonso, Ramón Rad Moradillo, Juan Carlos Cambra Baseca, Carlos Castroviejo Fernández, Mª Pilar Barros García, Rocío Herrero Cosío, Álvaro https://hdl.handle.net/10259/11727 2026-05-27T00:05:34Z 2025-10-01T00:00:00Z

SEM-EDS and hyperspectral images of vine leaves treated with antifungal products Sánchez Alonso, Ramón; Rad Moradillo, Juan Carlos; Cambra Baseca, Carlos; Castroviejo Fernández, Mª Pilar; Barros García, Rocío; Herrero Cosío, Álvaro Scanning electron microscope, better known by its acronym as SEM, is a very useful technique for obtaining highresolution images of the surface of a sample. Hyperspectral imaging provides precise information for analysing vineyard vegetation that could help in improving pesticide application in precision viticulture technics. The present dataset is based on images of vineyard leaves, taken with both technics. The leaves of the cv. Tempranillo, proceeding from a vineyard located inside of the Cigales Denomination of Origin, in north-central Spain, were treated with two Cu-containing products: ZZ Cuprocol (70 % w/v copper oxychloride) and Cuprantol Duo (14 % w/w copper oxychloride, 14 % w/w copper hydroxide). In addition, a contact pesticide widely used in intensive and traditional viticulture based on Folpet, copper-free but containing sulphur and chlorine, has been tested in its commercial form, Vitipec Blue (Cymoxanil 6 % w/w, Folpet 37.5 % w/w, Ascenza, PT). Three dilutions were prepared, one of each compound, at the actual field application concentration of 1.33 g/L. The leaves were sampled and processed during the 2023 season. These leaves were taken from the central part of representative shoots of the vine canopy, with east and west exposures. After the application of the pesticide dilutions, images of the leaves were taken with a 300-channel hyperspectral camera (Pika L, Resonon) using a mechanical bench synchronized with the camera. Then the SEM analysis was carried after prepare the samples. Hence, such imagery is provided in the present dataset, based on the images taken from the leaves with both technics

2025-10-01T00:00:00Z Rodríguez Villagrá, César Martin Reizabal, Sergio Ruiz González, Rubén Basurto Hornillos, Nuño Herrero Cosío, Álvaro https://hdl.handle.net/10259/11497 2026-04-08T11:34:34Z 2026-04-12T00:00:00Z

Labeled IoT Window-Based Random Network Pattern Dataset for Reinforcement Learning Rodríguez Villagrá, César; Martin Reizabal, Sergio; Ruiz González, Rubén; Basurto Hornillos, Nuño; Herrero Cosío, Álvaro This dataset is designed to support the training and evaluation of reinforcement learning models in the context of network traffic analysis. It is derived from an existing IoT network traffic dataset, from which packet capture (pcap) files were selected and processed following a custom methodology explained in [Methodological Information](methodological-information). The resulting data representation is based on a windowing approach, where network traffic is segmented into fixed-size temporal windows. Each window aggregates traffic instances and is labeled according to its composition as benign, attack, or mixed (containing both benign and malicious activity). The final datasets are generated through random combinations of these windows, enabling the creation of diverse traffic patterns that better reflect dynamic and random network conditions. This structure facilitates the use of the dataset in reinforcement learning scenarios, where agents must learn to identify, classify, or respond to varying traffic behaviors over time. Additionally, the evaluation datasets are generated following the same methodology as the training datasets, but are kept separate and are not used during the training process, allowing for an independent evaluation of model performance.

2026-04-12T00:00:00Z