Visible Light Communication (VLC) and Optical Camera Communication (OCC) are recent viable technologies for wireless communication, using existing lighting infrastructure and conventional cameras. The Color-Hopping Space-Time (CHST) scheme is a promising technique for VLC/OCC, introducing permutations in the mapping between symbols and colors. This work presents the first experimental validation of CHST in an OCC system, using an RGB LED array and a camera, with an accessible implementation based on low-cost hardware. The results confirm improvements in channel conditioning, BER reduction, and increased communication range with the use of the CHST scheme.
This study validates LoRaWAN coverage in forested and sloped terrains using field tests and simulations. Heltec LoRa modules operating at 433 and 915 MHz were evaluated under varying topography, vegetation density, and environmental interference. Experimental results show that adjusting LoRa parameters (spreading factor, bandwidth, and coding rate) significantly improves communication range, even in non-line-of-sight (NLoS) conditions. Terrain-aware simulations with Radio Mobile, based on the Longley–Rice model and SRTM data, correlated well with empirical results, but highlighted limitations in predicting dynamic factors such as multipath fading and vegetation absorption. This combined approach offers practical insights for optimizing LoRaWAN network deployments in challenging natural environments, making it suitable for environmental monitoring, remote sensing, and rural IoT applications.
The increasing frequency and intensity of natural disasters in Chile demands innovative technological solutions to strengthen the timely and efficient risk management capabilities of the National Disaster Prevention and Response System (SINAPRED).Based on the guidelines and coordination plans promoted by the National Disaster Prevention and Response Service (SENAPRED), this work proposes the design and implementation of a satellite platform focused on Earth observation and geospatial analysis, supported by quantum communication capabilities. BThrough the use of data from optical sensors, radars, and LoRa satellites, the platform will integrate near real-time processing capabilities and intuitive visualization tools for institutional and technical users. This technological integration will strengthen technological autonomy and significantly improve data processing efficiency. The focus of this paper is a model designed to achieve greater efficiency in the data processing workflow.
Integrated Sensing and Communication (ISAC) has emerged as a promising solution to spectrum scarcity and interference issues in next-generation wireless networks. In particular, resource integration enables the coexistence of sensing and communication functions using a single waveform within shared spectral resources. Despite the growing theoretical interest, empirical validation of ISAC systems, especially under practical conditions, remains limited. This paper presents a proof-of-concept implementation of a wireless ISAC system based on a bistatic sensing configuration and a target. The system employs a time-division multiplexed (TDM) waveform consisting of a linear chirp for sensing and a binary phase shift keying (BPSK)-modulated signal for communication. A hardware-reconfigurable platform was developed, and the system was validated through a real-world measurement campaign. The obtained experimental results confirm the feasibility of the proposed architecture and highlight its potential for practical ISAC deployments in complex environments.
This paper introduces ModernBERT-DoS, a transformer-based system designed for the detection and classification of network traffic, with a particular focus on identifying Distributed Denial of Service (DDoS) attacks. The model is fine-tuned on the CIC-DDoS2019 dataset and an additional custom SSL dataset, enabling robust performance across a wide range of traffic categories. Leveraging a modified BERT architecture, ModernBERT-DoS achieves high accuracy, precision, recall, and F1-scores in the evaluations. The training process incorporates early stopping to optimize generalization and prevent overfitting. For transparency and reproducibility, the trained model is publicly available via the Hugging Face Hub.
This paper presents a simulation framework for modeling visible light communication systems that employ color shift keying modulation and rolling shutter image sensors. Implemented as a modular Python package, the simulator integrates ray tracing for light propagation models of indoor environments, camera optics, color filtering, and noise. The rolling shutter mechanism is accurately represented, enabling the spatial and temporal analysis of color shift keying signal reception. Simulation results using a Cornell Box setup demonstrate the system’s capability to generate synthetic images that reflect the effects of transmission frequency, gain, and environmental geometry. Applications include the analysis of inter-channel interference, generation of datasets for machine learning-based decoding, and the design of optimized color shift keying constellations.
The increasing demand for efficient communication systems in the Internet of Things (IoT) era has motivated the exploration of Optical Camera Communication (OCC) as a cost-effective alternative to traditional Radio Frequency (RF) technologies. This study presents a practical implementation of a modulation scheme inspired by APSK-CSK, introducing a novel Hue and Saturation Shift Keying (HSSK) design based on the Hue Saturation Value (HSV) color space. The system enables reliable data transmission using a single RGB LED and a regular smartphone camera, targeting low-power and low-cost IoT applications where simplicity and robustness are prioritized over throughput. This implementation employs a single LED to validate the feasibility of the system and lays the foundation for future multitransmitter OCC systems. Using Hue-based circular constellations, the proposed approach simplifies the modulation process while achieving a reliable data rate of 100 bps at 5 meters with a BER of 10−3 under realistic lighting conditions. These results highlight the potential of HSSK-based OCC as a practical and scalable solution for IoT applications utilizing existing camera and lighting infrastructures. While this work demonstrates feasibility with an ESP32 and a consumer smartphone (Poco X3 Pro), higher data rates could be achieved with more powerful hardware. However, the advantage of this system lies in its use of standard cameras without losing their primary functionality.
With increasing demand for high-speed and interference-resilient wireless connectivity, LiFi emerges as a promising alternative to traditional RF-based technologies. This study experimentally evaluates the performance of a commercial LiFi system (LiFiMAX®) in a real residential setting. Two user load conditions, single user and dual user, were tested across a 3×3 spatial grid covering nine measurement points at three different heights (0 cm, 72 cm, and 145 cm). The experiments used ICMP-based ping tests and UDP throughput measurements using iperf3 to evaluate key performance indicators: download speed, upload speed, and latency. Under optimal conditions, download speeds reached up to 69 Mbps with latencies as low as 6.33 ms. When two users accessed the same optical channel simultaneously, throughput dropped by up to 42%, and latency increased by approximately 35%. These results highlight the spatial sensitivity of the system and the performance degradation under multi-user contention, providing insights into practical deployment strategies of LiFi networks in home environments.
WiFi-based sensing is a privacy-preserving approach to Human Activity Recognition (HAR), especially in restricted environments. While deep learning models using Channel State Information (CSI) have shown high accuracy, their lack of explainability limits deployment in sensitive applications. This study applies Grad-CAM and SHAP to a previous proposed HAR classification pipeline to better understand which signal features drive predictions. We evaluate model generalization across devices and environments, and analyze explanation consistency. Results show both methods can highlight relevant subcarriers, with SHAP offering more consistent insights at higher computational cost. Our findings underscore the need for more realistic datasets and suggest that combining CSI with other modalities may improve model transparency. To ensure the reproducibility of our work, all the developed code is open-source.
This work presents a clustering-based methodology for IP block allocation in Networks-on-Chip (NoC), leveraging a multipartitioning extension of the Kernighan–Lin algorithm (MKL). By minimizing cut cost—interpreted as inter-cluster communication—MKL enhances locality and reduces global interconnect usage. Unlike recursive bipartitioning, MKL operates directly on k-way partitions, aligning with SRMI (Single-Router Multiple-IP) architectures. Experiments on standard NoC benchmarks confirm that MKL consistently improves intra-cluster traffic and reduces communication overhead.
Considering the 2028 IEEE 802.11bn physical layer (PHY) that schedules modulation code schemes (MCS) with either 4096 quadrature amplitude modulation (4K-QAM) or 16K-QAM signaling schemes impaired with phase noise (PN) and carrier frequency offset (CFO), this paper analyses the performance of Chase Combining Hybrid Automatic Repeat Request (CC-HARQ) protocol with Low-Density Parity-Check (LDPC) codes using the following transceiver architecture: transmit beamforming singular value decomposition precoder (SVD) and minimum mean squared error (MMSE) detector. Simulation results show that standard schemes can mitigate the power losses due to CFO, while the system performance is dramatically affected by the joint effects of PN and CFO. Fortunately, the implementation of CC-HARQ protocol can reduce these power losses due to PN and CFO hardware impairments, e.g., from 6.4 dB to 1.6 dB for MCS13 (4K-QAM, LDPC code with code rate 5/6), and from 7.6 dB to 2.0 dB for MCS15 (16K-QAM, LDPC code with code rate 5/6), assuming only two retransmissions. Notably, as quantified in this paper, these performance enhancements due to CC-HARQ protocol depend strongly on PN power, channel model (block fading or independent), and the time scale of channel state information (CSI) feedback mechanism.
The Internet of Things (IoT) is based on data collection for future processing and decision-making. In multihop Low-Power and Lossy Network (LLN) scenarios, it is important to collect data from sensors and IoT devices with efficient energy consumption. This paper explores the concept of multi-mobile agents (MMA) combined with proactive caching to collect data in an IoT network. In the proposed architecture, the group of itineraries is computed by a mechanism that uses the Hamming distance technique to partition a sensor network into a number of data-centric clusters. In addition, the data collection mechanism is modeled as a knapsack problem that improves the energy efficiency with MA payload variation. The simulations show lower energy consumption from network devices as the payload and the number of MMA increase.
This work summarizes an integrated strategy for underground optical communications combining technology selection and advanced processing. First, an affine transformation T(x) = Ax + b compacts and nearly isotropizes the data clouds, aligning orthogonal decision boundaries and simplifying threshold detection. Second, the integration of an S2-CNN with softmax as soft metrics for Ring-TCM reduces the BER by 35–45% under NLoS conditions compared to classical decoding. Under channel conditions, VLC+CSS excels in clean LoS (BER ≤ 10−5 with Eb/N0 ≈ 7 dB), while IR+CSS dominates severe NLoS due to its greater robustness to scattering. Together, the affine+neural pipeline decreases implementation complexity and increases resilience, enabling reliable links in mining tunnels with varying dust and humidity.
The rapid growth of the Industrial Internet of Things (IIoT) poses significant security challenges due to constrained resources in devices. Traditional threat detection approaches are inadequate for such environments because they do not consider the low capacity of the devices in a scenario where the detection must be done in a short time interval. This paper proposes a methodology that combines an analysis of sufficient training sample sizes with automatic feature selection techniques of the protocol information for effective anomaly detection. Experiments using an expanded BRURIIoT dataset demonstrate that this approach significantly reduces computational resources and network traffic without compromising detection accuracy, offering an efficient solution for secure IIoT networks. For instance, we observed that by removing 84.5% of the original dataset and 74.9% of the original features, the accuracy was reduced by only 0.16%. Further, with the efficient improvements in the dataset, it was possible to run the inference in constrained devices, something not possible with the original dataset, detecting some attacks in less than 6 seconds. Following sound open science practices, all developed code is shared as free and open-source software.
This work presents an algebraic reformulation of the Kernighan–Lin (KL) algorithm for graph partitioning, using broadcast operators and tensorial structures. This unified formulation allows the algorithm to be extended from the bipartite case (k = 2) to the multipartite case (k > 2) in a compact, scalable, and vectorizable way. The proposed Multipartitioning KL (MKL) algorithm is expressed in terms of Boradcast Algebra and its performance validated through 15 representative traffic benchmarks in Network-on-Chip (NoC) architectures. Results show a consistent reduction in cut cost and a favorable redistribution of traffic — more intra-cluster and less inter-cluster — which is especially relevant in Ad-Hoc NoCs communication topologies. This formulation positions MKL as an effective and architecturally relevant heuristic, while establishing an algebraically robust foundation for its extension to more complex scenarios, such as dynamic partitioning or regular architectures.
Industrial systems are increasingly interconnected, creating a complex landscape that exposes critical infrastructure to advanced cyber threats. Traditional security measures often fail to address the vulnerabilities in industrial environments where both Information Technology (IT) and Operational Technology (OT) components must be secured. This paper proposes a novel security architecture that integrates blockchain-based Self-Sovereign Identity (SSI) with multi-layered endpoint protection and leverages the principles of the National Institute of Standards and Technology’s (NIST) Zero Trust Architecture (ZTA). By using blockchain-based SSI, the proposed framework provides decentralized and immutable identity verification for all devices, users, and applications and ensures that only authenticated entities can interact within the industrial network. This architecture includes a multi-layered defense model that enables contextual access controls in all layers, from the information systems layers to the control network layer. We evaluate the performance of the architecture, in particular, the latency associated with identity verification and revocation within the critical layers, to show that blockchain-based SSI can enhance security without compromising operational efficiency.
A comprehensive framework for hyperparameter optimization is developed in neural receiver models for Orthogonal Frequency Division Multiplexing (OFDM) wireless communication systems using the Estimation of Bayesian Network Algorithm (EBNA). The proposed methodology addresses the critical challenge of selecting optimal hyperparameters for convolutional neural network-based receivers in 5G OFDM systems. Through the application of EBNA, an algorithm that constructs probabilistic models to capture hyperparameter interdependencies, we perform a systematic optimization of both the learning rate and the number of residual blocks, with the objective of maximizing the Bit-Metric Decoding (BMD) rate. Our experimental results demonstrate that the optimized neural receiver achieves performance comparable to baseline models while requiring only 5% of the original training iterations, suggesting the existence of local optima that can be efficiently identified through Bayesian optimization. The hierarchical framework developed provides a scalable approach for future hyperparameter exploration in neural receiver communication systems.
Detecting anomalies in high-dimensional or unstructured data remains a challenge, as many traditional approaches such as Autoencoders (AEs), Variational Autoencoders (VAEs), and One-Class SVMs rely heavily on reconstruction errors or distance measures. These techniques often overlook contextual information, struggle with subtle deviations, and provide limited interpretability. To address these gaps, we propose AE-RAGX, a hybrid architecture that combines Autoencoders with Retrieval-Augmented Generation (RAG) using Large Language Models (LLMs). By transforming structured data into descriptive text and incorporating rule-based knowledge, the model not only improves anomaly detection but also produces transparent, human-readable explanations. Evaluation on a credit card fraud dataset shows that AE-RAGX enhances recall and provides interpretable outputs, making it well suited for high-stakes applications such as fraud detection.
This paper presents a simulation-based parametric analysis of an LED-based optical wireless positioning (OWP) system integrated with YOLOv12 for drone detection under adverse atmospheric conditions. The study evaluates 18 LED configurations varying semi-angle (60°–90°), power (1–50 W), and density (15 vs. 45 LEDs) across five visibility conditions (0.5–20 km) representing fog and smoke scenarios in a 15×25×5 m–room. YOLOv12 models trained on 1440 images with different configurations (25/50 epochs, with/without cross-validation) achieved up to 0.992 mAP50 and 0.981 precision. LED positioning analysis across 33000 simulated samples per configuration shows exponential error reduction from 15 cm at 1 W to 0.3 cm at 50 W, validating joint optical–visual optimization.
This document addresses the problem of indoor positioning, where the Global Positioning System (GPS) presents limitations such as interference, spectrum saturation, and low prediction capacity. As an alternative, a Visible Light Positioning (VLP) system is proposed. An Extreme Learning Machine (ELM) neural network is used to predict the location (x,y) of a receiver (photodiode) in a database constructed in a controlled environment with 7344 samples containing RSS values. The model was trained using 5-fold cross-validation and evaluated with metrics such as root mean square error (RMSE), correlation coefficient (r), and training time. The ELM results achieved an accuracy of 13.50 cm (RMSE) using 200 neurons, with a value of r=0.9971 and a training time of ms, reinforcing its applicability in real-time. Additionally, the results are compared with state-of-the-art models that constructed the database, where the best performance is achieved by the ridge regression approach, with an accuracy of 13.3 cm (RMSE). Therefore, the ELM not only achieves comparable performance but also offers greater computational efficiency, positioning it as a feasible proposal for indoor localization systems.
