Resource
2026 EN
Ioanna Mandilara · Christina Maria Androna · Eleni Fotopoulou
+19 more
Several data and knowledge management services have recently been produced to address the challenges related to climate change. However, many of the proposed solutions are not made openly available or consider their operation over a set of assumptions in terms of data usage, limiting the potential for their adoption by a wider community. Furthermore, most of the existingwork is operational in the form of software or data silos, without giving high importance to interoperability and extensibility features and with a high level of obscurity in the supported mechanisms. This article details the Knowledge Management Framework developed in the ARSINOE project, which is an EU-funded project aimed at creating climate resilient-regions through systemic solutions and innovations. The ARSINOE Knowledge Management Framework leverages modern technologies such as Knowledge Graphs and Digital Twins. It includes a Data Hub to host and make available heterogeneous environmental and climate data, a Knowledge Graph that tracks information related to the Sustainable Development Goals and enables the development of participatory socio-ecological modeling and analysis, interfaces for the development of Digital Twins that onboard socio-ecological models, as well as visualization and analysis services. The ARSINOE Knowledge Management Framework is open, modular and extensible by design, while it addresses data quality assessment and management challenges, strengthening its adoption, comparability, and replicability in different locations. The general framework is presented, focusing on the functionality of each component, their interplay, and its applicability in specific case studies throughout Europe.
Resource
2026 EN
Jameel Ahmad · Usama Ahmed · Saad Farooq
+4 more
Distributed Denial of Service (DDoS) attacks in the constantly expanding sphere of cybersecurity are highly important to detect and respond to promptly. This paper proposes an Intrusion Detection System (IDS) that integrate Machine Learning (ML), Apache Kafka and Apache Spark to enable real-time detection and mitigation of DDoS attacks. The proposed framework collects live network traffic data, performs preprocessing and feature extraction, and applies machine learning algorithms within a streaming architecture supported by Apache Kafka and Apache Spark. Kafka enables real-time data ingestion, while Spark provides distributed processing for efficient model training and inference. The methodology supports adaptability to emerging attack patterns through training and evaluation on publicly available benchmark datasets.Machine learning models, including Random Forest and Support Vector Machine (SVM), were evaluated within the proposed streaming framework. Experimental results show that the integrated use of Apache Kafka and Apache Spark achieved an accuracy of up to 95%. The real-time streaming capability of Kafka, combined with the distributed processing power of Spark, enables efficient handling of high-volume network traffic and supports timely detection of DDoS attacks.
Resource
2026 EN
Md Sharif Hossen · Cole Dickerson · Ozgur Ozdemir
+33 more
In this work, we present an unmanned aerial vehicle (UAV) wireless dataset collected as part of the AERPAW Autonomous Aerial Data Mule (AADM) challenge, organized by the NSF Aerial Experimentation and Research Platform for Advanced Wireless (AERPAW) project. The AADM challenge was the second competition in which an autonomous UAV acted as a data mule, where the UAV downloaded data from multiple base stations (BSs) in a dynamic wireless environment. Participating teams designed flight control and decision-making algorithms for selecting which BSs to communicate with and how to plan flight trajectories to maximize data download within a mission completion time. The competition was conducted in two stages: Stage 1 involved development and experimentation using a digital twin (DT) environment, and in Stage 2, the final test run was conducted on the outdoor testbed. The total score for each team was compiled from both stages. The resulting dataset includes link quality and data download measurements, both in DT and physical environments. Along with the USRP measurements used in the contest, the dataset also includes UAV telemetry, Keysight RF sensor position estimates, link quality measurements from LoRa receivers, and Fortem radar measurements. It supports reproducible research on autonomous UAV networking, multi-cell association and scheduling, air-to-ground propagation modeling, DT-to-real-world transfer learning, and integrated sensing and communication, which serves as a benchmark for future autonomous wireless experimentation.
Resource
2026 EN
Marta Iovino · Ivan Lazic · Chiara Bara
+3 more
This study introduces kCMI-FS, a feature selection (FS) method that leverages Conditional Mutual Information (CMI) estimated via an adapted k-nearest neighbour (kNN) strategy to handle mixed-type data with continuous features and discrete targets. Unlike traditional approaches, based on Mutual Information, that may overlook redundancy or higher-order dependencies, kCMI-FS incorporates a significance-based forward selection process to identify informative and non-redundant features. We assess its performance on theoretical simulations, five synthetic datasets, and four biomedical benchmark datasets that highlight key FS challenges. Results demonstrate that kCMI-FS consistently recovers relevant features in structured scenarios and matches or outperforms existing methods, particularly in mixed-variable and high-dimensional conditions, even if in some cases at the price of a few more redundant/irrelevant features selected. Furthermore, classification experiments carried out on the biomedical datasets confirm that kCMI-FS offers strong predictive performance with reduced feature sets, thus enhancing model interpretability without compromising accuracy compared to existing methods. The results highlight the potential relevance of kCMI-FS in biomedical data analysis, particularly in classification problems where interpretability, feature compactness, and robustness are essential for supporting early diagnosis and clinical decision-making.
Resource
2026 EN
Lamprini Damakoudi · Stelios Simos · Dimosthenis Spasopoulos
+11 more
This work presents an integrated plasmophotonic interferometric sensor providing environmental temperature and thermo-optic coefficient (TOC) measurement capabilities. By exploiting extreme light field concentration on the plasmonic metal stripe, the sensor employs a liquid dielectric layer deposited on the plasmonic sensing area for refractive index changes versus temperature. Experimental characterization was conducted with four different liquids, achieving a record temperature sensitivity of +23.86 nm/°C for a 70% ethanol–water solution. Additionally, by leveraging the sensor’s high sensitivity, the TOC of an unknown liquid was determined with a relative error of 13.1%.
Resource
2026 EN
Thanet Markchom · Olympia Kourounioti · Matteo Marturini
+14 more
Maritime surveillance plays a vital role in protecting coastal and maritime environments. However, traditional maritime surveillance systems that rely on singlealtitude, single-modality sensors suffer from limited coverage and sensitivity to weather conditions. To address these limitations, this article presents a comprehensive maritime surveillance system that integrates multialtitude, multimodal sensor platforms, including ground-based sensors, lowaltitude uncrewed aerial vehicles (UAVs), and satellites, for maritime threat detection. Each platform is equipped with dedicated modules for object detection, tracking, and geolocation, leveraging its unique sensing capabilities to contribute to a coordinated surveillance system. Moreover, a novel multialtitude, multimodal maritime surveillance (MAMMS) dataset is introduced. This dataset includes data from these sensor types, enabling rigorous benchmarking across varying operational conditions. The experimental results indicate that the system achieved an average mAP of 50.5% across all sensors in object detection, surpassing state-of-the-art models in most cases. For object tracking, the system achieved an average ID F1-Score (IDF1) of 0.263 and a higher order tracking accuracy (HOTA) of 0.297, comparable to state-of-the-art methods, while exhibiting substantially fewer average ID switches (IDSWs) (75.46) compared to the strongest baseline (301.46). For geolocation approximation, the system achieved an error of less than 11 m in certain scenarios. A case study was also conducted to assess the sensor platforms when integrated into a multisensor fusion system. The case study showed that complementary information from different platforms can help reduce false alarms and improve object geolocation accuracy. The dataset is available at https://zenodo.org/records/17979190
Resource
2026 EN
Nikos Temenos · Anastasios Temenos · Charalampos Zafeiropoulos
+5 more
Deep Learning (DL) models have demonstrated remarkable performance in Remote Sensing (RS) land use land cover (LULC) classification. Yet, their high computational complexity demands can limit their deployment in resource-constrained edge computing environments. Existing structured pruning approaches are often weight-dependent and lack a principled mechanism to jointly coordinate the pruning process and select the optimal network sparsity ratio. To address these limitations, this work introduces PrONe-RS, a graph-isomorphic pruning coupled with Pareto-driven optimization framework for efficient neural network sizing in RS LULC applications. Through the incorporation of Isomorphic Structured Pruning (ISP), the network's structure is modeled as a graph to identify and create isomorphic groups of structurally similar computational blocks, which are consequently pruned according to a user-defined ratio. With this principally coordinated pruning process across layers, ISP preserves the network's structural coherence making it attractive for DL models utilizing diverse computing blocks. PrONe-RS is enhanced with a Pareto-driven multi-objective optimization scheme that first identifies the optimal pruning ratios achieving the best classification performance and complexity trade-off, and then automatically selects the most efficient one through a dedicated metric, thereby guiding the pruning ratio selection process. Extensive ablation study on the effective pruning values and the selection of the optimal one balancing the performance-complexity trade-off using five widely-used DL models including EfficientNet, ResNet, ResNeXt, VGG and Vision Transformer variations across two datasets, EuroSAT-Multispectral and AID, demonstrate significant reduction in number of trainable parameters, corresponding to $64-98\%$ for the EuroSAT-Multispectral case and $44-70\%$ for the AID case, all while preserving $>91\%$ classification performance across all metrics. Comparisons with existing pruning approaches on AID dataset and EuroSAT-RGB, demonstrate improved classification accuracy along with a substantial reduction in parameters, achieving a 91% reduction for EuroSAT-RGB, 36% for ResNet, and 75% for VGG on AID, respectively.
Resource
2026 EN
Daniel Roggen · Megan Walker · Yojan Patel
+26 more
The Pixel Watch 2 (PW2) is the first Google smartwatch to combine multipath photoplethysmography (PPG) with deep learning-based heart rate inference, designed to significantly improve sensing accuracy during motion-heavy activities. The device processes 10 optical channels using an on-device, 15-layer temporally dilated convolutional neural network ( $\sim$ 300K parameters) to yield a 1 Hz heart rate output. Crucial to this model's performance was its training on a massive dataset comprising 10,000 hours of data from 962 participants, curated from a broader corpus of controlled and free-living activities. We evaluated the PW2's sensing performance across two independent validation sets: an in-house fitness dataset (229 participants, 250 hours) and an external free-living dataset (27 participants, 1000+ hours). The system achieved 95% Limits of Agreement of -10.34 to 8.66 BPM during exercise and -6.57 to 7.48 BPM during free-living activities, demonstrating substantially tighter error margins than previous Google devices. Finally, we discuss key design lessons, emphasizing that large-scale deep learning was instrumental in fully leveraging multipath PPG hardware over traditional signal processing approaches.
Resource
2026 EN
Dimitris E. Simos · Manuel Leithner · Rick Kuhn
+3 more
Ten years ago, we introduced a research program applying high strength combinatorial test methods to vulnerability and fault detection for cybersecurity. This article discusses advances in the field and where these methods may be effective for today’s complex security problems.
Resource
2026 EN
Jyoti Prakash Sahoo · Binayak Kar · Yi-Leh Wu
+2 more
The widespread adoption of Industrial Internet of Things (IIoT) devices at the network edge necessitates advanced intrusion detection systems to mitigate sophisticated cyber threats. Edge IIoT environments, characterized by heterogeneous data and resource constraints, demand adaptive, scalable, and interpretable security solutions. This study proposes a sparse mixture of Quantum Kolmogorov-Arnold Experts Intrusion Detection system (SMoQKE-IDS), an innovative framework that integrates quantum-classical machine learning, federated learning, and interpretable AI within a Mixture of Experts (MoE) architecture. Unlike existing quantum-enhanced or Kolmogorov-Arnold Networks (KAN)-based IDS, which typically rely on standalone architectures or dense, unified ensembles, SMoQKE-IDS introduces a Sparse Mixture of Experts (SMoE) paradigm to decouple model capacity from computational overhead. The framework integrates parameterized quantum circuits for high-dimensional Hilbert-space feature mapping, convolutional layers for spatiotemporal feature extraction, and KANs for interpretable univariate nonlinear modeling. These heterogeneous modules are governed by a modality-specific sparse gating mechanism that implements conditional computation by dynamically activating the top-k optimal expert sub-networks for each input. This approach effectively mitigates the computational redundancy inherent in dense models while enhancing adaptability to diverse network traffic profiles. Additionally, federated learning enables privacy-preserving and scalable training, while the inherent interpretability of the framework enhances trust through visual and analytical transparency. Rigorous validation on benchmark datasets within federated environments demonstrates superior detection accuracy, heightened transparency, and robust performance, establishing SMoQKE-IDS as a robust solution for securing edge IIoT ecosystems.