Journals
2026 EN
Dibbern Thais · Salles-Filho Sergio · Romani Luciana Alvim Santos
+1 more
This paper aims to investigate the outcomes of scientific production linked to the Brazilian Agricultural Research Corporation (Embrapa) in the policy-making process, addressing key issues such as the type of use of this expertise, the leading national and international institutions that use it, and the predominant topics addressed in these documents. Methodologically, an exploratory study was conducted using bibliographic and documentary review activities, as well as the Overton data source. The results highlight Embrapa’s relevance not only at the national level but also in international contexts, indicating its substantial influence on policy documents produced by governmental and intergovernmental bodies, especially in the Global North—a result linked to the limitations of the database used. The assessment of topics associated with the identified policy documents indicates that the knowledge and technologies produced by Embrapa extend beyond the agricultural sector, encompassing other areas such as “sustainability and climate change”, “deforestation and forest degradation”, and “health and poverty”. This case study serves as a concrete application of emerging scientometric methods for tracing research impact beyond academia. It demonstrates the potential of policy document analysis to capture the broader societal contribution of mission-oriented research institutions, while also critically examining the methodological limitations of such tools for global research evaluation. The analysis highlighted themes of outcomes other than the traditional competencies of Embrapa, notably new cultivars, genetics, and agronomic techniques. The potentialities and limitations associated with the Overton database were also considered, taking into account the results obtained.
Resource
2026 EN
Junhyeong Lee · Hyun Kwon
Malicious Uniform Resource Locator (URL) detection remains a critical and actively researched topic in the field of cybersecurity due to the increasing prevalence and sophistication of web-based threats. In modern military communications, malicious URLs can disrupt command-and-control systems, leading to compromised operational security. Consequently, robust URL detection in military networks is paramount for maintaining mission continuity. However, existing single-model approaches often fall short in addressing the diverse structural and semantic characteristics of modern malicious URLs. To overcome these limitations, this study proposes a hybrid ensemble detection framework that integrates tree-based models with a Bidirectional Encoder Representations from Transformers (BERT). In the data preprocessing stage, key structural features were extracted from URLs, and tokenized sequences were prepared for BERT input. Tree-based models were independently trained on the extracted features, while the BERT model was fine-tuned for binary classification of malicious URLs. Hyperparameters for the tree-based models were optimized using Optuna, and a sampling strategy was adopted for BERT training to mitigate class imbalance and computational cost. Soft Voting was applied to the tree-based models to enhance their collective performance, and the final predictions were generated through Weighted Voting that combined outputs from both the tree-based ensemble and the BERT model. Experimental results demonstrate that the proposed hybrid ensemble significantly outperforms traditional single-model baselines and simple ensemble methods, achieving improved detection accuracy and robustness. These findings demonstrate the effectiveness and practical applicability of the proposed hybrid approach in real-world malicious URL detection systems.
Resource
2026 EN
Gu Yongzhen · Wang Mengtian · Wang Haoxin
Conventional space-borne antenna shaping designs predominantly rely on a pre-defined reflector geometry, which lacks the capability to adapt to varying multi-stage service requirements once in orbit. To overcome this limitation, this paper proposes an active shaping method for space-borne membrane reflector antennas (MRA) through coordinated control of electrostatic forces and boundary cable tensions. A coupled structural-electromagnetic model is established that integrates the nonlinear finite element method for structural mechanics and the physical optics approach for radiation analysis, capturing the interactions between actuation forces, reflector morphology, and far-field performance. With electrostatic voltages and cable tensions as design variables, an optimization model is formulated to minimize the root mean square error of the directivity at sample points within the target coverage region. Numerical simulations conducted on a 30-meter aperture MRA demonstrate the effectiveness of the method, achieving an average directivity improvement of approximately 2.67 times after active shaping. The approach provides a feasible strategy for in-orbit reconfiguration of membrane reflector antennas to accommodate dynamic mission profiles and was verified through experiments with 5-meter electrostatically formed MRA.
Resource
2026 EN
Bashir Olaniyi Sadiq · Mohammed Dahiru Buhari · Yale Ibrahim Danjuma
+1 more
Flying Ad Hoc Networks (FANETs) face significant challenges in maintaining reliable peer-to-peer (P2P) communication due to high node mobility, dynamic network topologies, and uncertain link quality, which collectively degrade throughput, energy efficiency, and fairness. To address this problem, this paper proposes a Smell Agent Optimization-based Peer-to-Peer Selection (SAO-P2PS) algorithm designed to optimize peer selection under realistic channel uncertainties. The methodology uses a bio-inspired SAO framework that dynamically adapts to fluctuating link conditions by modeling peer selection as a multi-objective optimization problem, simultaneously maximizing throughput and energy efficiency while ensuring equitable resource distribution using fairness index. The algorithm incorporates uncertain link quality estimates directly into its utility function, enabling robust decision-making in highly dynamic environments. Simulation results show that SAO-P2PS significantly outperforms conventional Random Selection and Nearest-Neighbor Selection methods, achieving 25-35% higher throughput, 40-60% better energy efficiency, and maintaining fairness indices above 0.6 across varying link quality conditions. The proposed algorithm exhibits rapid convergence within 20-30 iterations and maintains stable performance despite UAV mobility and channel fluctuations. The significance of these findings lies in providing a computationally efficient, adaptive solution for real-time FANET applications where reliable P2P communication is critical for mission success. This research establishes SAO-P2PS as a foundation for intelligent peer selection strategies in next-generation aerial networks, with potential applications in disaster response, surveillance, and autonomous swarm operations in 5G-and-beyond communication ecosystems.
Resource
2026 EN
Mojtaba Namvar · Ebrahim Amiri · Mahdi Jafari-Nadoushan
Coverage evaluation in the orbital design of satellite constellations is a major driver of project cost because it dictates the required number of satellites. This becomes particularly critical when the objective is continuous 24/7 coverage, which requires eliminating coverage gaps while maintaining an optimal satellite count. Numerous methods have been proposed for evaluating coverage, each involving trade-offs between accuracy and computational load. Considering the limitations of previous methods, this paper presents a critical-point-based approach that focuses on spatial boundary locations and footprint intersections where coverage loss within the target area is most likely. The method improves coverage-evaluation accuracy while reducing computational complexity and sensitivity to spatial and temporal discretization. After presenting the underlying theory, we assess performance through case studies. We define mission parameters and perform orbit design via numerical optimization using particle swarm optimization (PSO) under simultaneous n-fold coverage constraints. The resulting designs are re-evaluated with the conventional grid point approach (GPA), yielding consistent outcomes. Although the numerical evaluation in this paper focuses on Walker-Delta orbital patterns and conical sensor footprints, the proposed critical-point framework is conceptually general and can be applied to other orbit families and footprint geometries.
Resource
2026 EN
Myoung-Ho Chae · Jin-Tae Park
Encrypted wireless traffic is difficult to analyze because link/physical-layer encodings, unknown frame structures, and payload cryptography hide higher-layer semantics. Yet accurate traffic classification under encryption is critical for spectrum situational awareness, QoS management, and electronic-warfare decision making. This paper studies blind service- and application-level traffic classification directly from demodulated L1 bitstreams, without protocol parsers, metadata, or cryptographic keys, in an over-the-air RF setting where a passive sensor observes only demodulated bits. We construct a MIL-STD-188-220–based wireless-network simulator that encapsulates encrypted application traces from the ISCX-VPN 2016 dataset and then applies cascaded link/physical-layer coding, interleaving, and scrambling to produce labeled demodulated bitstreams. The resulting corpus covers 6 service classes and 15 application classes under diverse encoding configurations and bit-error rates. On this corpus, we propose a BERT+BiLSTM classifier operating on bigram-tokenized bit sequences and compare it against several standard sequence backbones under a unified training setup. Experiments show that, even after two-stage encoding and at a bit-error rate of 10 –3 , classifiers that use only demodulated bitstreams achieve 94.03% service-level and 97.37% application-level accuracy, with graceful degradation at higher error rates. These results indicate that cascaded encoding statistically dilutes but does not erase class-discriminative structure, and that blind L1-bitstream traffic classification is a viable building block for future electronic-warfare capabilities such as platform/system discrimination, mission-state inference, and jamming-priority decisions.
Resource
2026 EN
Sonia Zappia · Ivan Iudice · Domenico Pascarella
+1 more
Reconfigurable Intelligent Surfaces (RIS) integrated with Unmanned Aerial Vehicles (UAVs) have emerged as a promising technology in the evolution of next-generation wireless communication systems. Herein, the potential of RIS-assisted UAV architectures are discussed, focusing particularly on their role in supporting Backscatter Communication (BackCom)— a low-power paradigm suitable for passive IoT devices — which remains little explored. To fill this gap, two representative paradigms are introduced: (i) UAV-empowered RIS-enabled BackCom and (ii) UAV-assisted RIS-enabled BackCom. These use cases are analyzed through a structured framework based on Concepts of Operation (ConOps), mission-level attributes, and Key Performance Indicators (KPIs), offering a comprehensive assessment of their performance, scalability, and deployment feasibility. In addition, a theoretical throughput-oriented performance analysis is developed, deriving closed-form rate (R) and signal-to-noise ratio (SNR) scaling laws as a function of the number of RIS elements and UAV transmit power. The analytical results highlight the quadratic SNR scaling behavior under ideal conditions and clarify the fundamental trade-offs between RIS aperture size and power beacon requirements. The findings demonstrate that UAV–RIS–BackCom architectures can achieve scalable, low-power, and rapidly deployable connectivity in infrastructure-limited environments, while also revealing practical limitations related to hardware impairments, control overhead, and channel dynamics. This work provides a unified conceptual and analytical framework to guide the design of sustainable, adaptive, and high-efficiency wireless systems toward practical 6G implementations.
Resource
2026 EN
Hossein Abroshan · Syed Waquas Hashmi
Backdoor attacks represent a serious challenge to robust deployment of machine learning (ML) and deep learning (DL) models in safety- and mission-critical fields. In a backdoor attack, an adversary injects a hidden trigger so that the model behaves normally when the inputs are clean but consistently produces attacker-chosen outputs when the trigger is present. Existing defences generally work at a single stage in the ML lifecycle-on data, on the model, or at inference time-and are thus susceptible to adaptive attackers that intentionally evade their underlying assumptions. This paper proposes Multi-Stage Backdoor Detection (MSBD) , which provides a defence-in-depth structure that combines multiple mechanisms in the training, post-training model inspection, and deployment-time monitoring. MSBD has four stages: influence-based screening of training samples (Stage A), optimisation-based trigger inversion (Stage B), neuron activation graph analysis for detection of suspicious subnetworks (Stage C), as well as calibrated runtime detection with integration of trigger signatures and perturbation-based consistency checks (Stage D). The platform is intended to function under realistic defender conditions with limited access to both data and models, and for offline validation and online monitoring. We evaluate MSBD on three benchmark datasets (MNIST, CIFAR-10, GTSRB) under a strong BadNets-style backdoor attack and compare it against five representative defences (STRIP, Neural Cleanse, Activation Clustering, Spectral Signatures, and Fine-Pruning). Across all datasets, we find an average F1-score of 99.0% for MSBD, which is consistently better than STRIP’s, with a practical runtime overhead, showing that multi-stage, cross-layer defences can significantly improve robustness over single-stage defences.
Resource
2026 EN
Potchakorn Poomekum · Thanapat Sorsoontornnirat · Arisara Suriyawong
+2 more
Large Language Model (LLM)–integrated enterprise applications introduce a dynamic and evolving attack surface, where vulnerabilities arise not from static code defects but from semantic misalignment, unsafe tool invocation, and multi-turn contextual reasoning. Existing security practices—including manual red teaming, prompt fuzzing, heuristic guardrails, and adversarial prompt libraries—remain limited in scalability and depth: they are labor-intensive, predominantly short-horizon, and rarely uncover multi-vector or multi-step exploit chains, while also lacking reproducible forensic evidence and systematic mitigation support. These limitations complicate the rigorous validation and hardening of complex LLM–tool–Retrieval Augmented Generation (RAG) systems. This paper presents Chimera-RL , an end-to-end autonomous red-teaming framework that operationalizes established reinforcement learning and attack-graph concepts within a unified system tailored to LLM-integrated applications. Rather than introducing a new learning algorithm, Chimera-RL combines curriculum-driven Hierarchical Reinforcement Learning (HRL), parameterized macro-actions, and mission-graph–guided exploration to enable long-horizon, campaign-level vulnerability discovery. The framework further integrates cryptographically hash-chained forensic logging and a template-driven mitigation playbook generator, providing verifiable attack traces and actionable defensive guidance.Experimental results demonstrate that Chimera-RL converges faster and more stably than standard deep RL baselines, including Proximal Policy Optimization (PPO) and Advantage Actor–Critic (A2C) , while discovering shorter and lower-cost exploit paths. When the generated mitigation playbooks are deployed, the attack success rate is reduced by nearly half, the median steps-to-compromise increase by a factor of seven, and the attacker’s operational cost more than doubles. These results indicate that Chimera-RL provides a practical and scalable system-level approach for automated red teaming and security hardening of LLM-integrated systems.
Resource
2026 EN
George J. Frangou
This paper presents the quantitative validation of Artificial Precognition Adaptive Cognised Control (APACC), a dual-layer neuro-symbolic architecture for safety-critical autonomous transport. APACC integrates Type-2 fuzzy symbolic reasoning for high-level decision-making with linearised predictive optimisation for trajectory control, synchronised through Diophantine Frequency Synthesis over a 2.5 s precognition horizon. We establish a theoretical framework that shows how these components interact to ensure bounded uncertainty propagation and temporal coherence across layers. Validation combines high-fidelity simulation with real-world field deployment. Automotive trials using Honda Civic vehicle dynamics achieved 51% reduction in peak deceleration (0.45 g to 0.22 g) and 67% decrease in jerk during pedestrian-crossing scenarios ( p < 0.01) compared with reactive baselines. Operational railway deployment under UK SBRI Project Edge on 32 km of live Network Rail infrastructure achieved 96% base-station-handover prediction accuracy, 18.6% latency reduction, and complete route coverage through predictive multi-modal connectivity management across four mobile-network operators with satellite backup-Long-short-term-memory (LSTM) signal-strength prediction achieved 5-8% RSRP error, confirming simulation-to-reality transfer. The system operated continuously for 168 h without failure, demonstrating robustness for mission-critical control. Across more than 10,000 simulation scenarios, APACC outperformed proportional-integral-derivative, model-predictive-control, and deep-reinforcement-learning baselines while maintaining interpretability consistent with ISO 26262 certification. All validation datasets, simulation environments, and operational telemetry are released via Zenodo for independent verification and reproducibility.