We propose a compact LIBS system with micro-spectrometers and optical filters, achieving a reduction in size and cost for water heavy metal detection while enhancing sensitivity and portability.
5G network slicing supports heterogeneous services (eMBB, URLLC, mMTC) over shared infrastructure but expands the attack surface to DDoS threats that violate SLAs. This paper proposes a deep learning-based DDoS detection framework using 1D-CNN. A realistic 5G standalone testbed (Free5GC, UERANSIM) generates benign and attack traffic (TCP/UDP/ICMP floods) across slices. Flows are extracted via CICFlowMeter and processed with sliding windows for temporal dependencies. Evaluated against LSTM and SVM baselines using Accuracy, Precision, Recall, F1-score, AUC, and MSE, the 1DCNN achieves 93.56% accuracy and 98.19% recall—outperforming LSTM $(79.68 \%)$ and SVM $(84.24 \%)$—with low inference latency ($0.0629 ~\text{ms} /$ sample), enabling real-time edge deployment. Results affirm CNN efficacy for robust DDoS detection in sliced 5G networks.
Application, after every grazing cycle, of 40 and 80 kg of nitrogen per hectare of sheep pasture, respectively, raised pasture yields to several times the level achieved in the unfertilized control. The fertilized herbage contained larger quantities of nitrogen compounds, with a trend to further increase in the following years. In variant A (40 kg N/ha) conversion of fertilizer nitrogen into plant matter was more efficient. In variant B (80 kg N/ha) we found a better balance of herbage regrowth over the whole grazing season. The amount of animal products produced in the individual variants correlated with the quantity and quality of the plant matter obtained.
With accelerating digital transformation, information system security is becoming a mandatory strategy. The surge in cyber threats, both in frequency and sophistication, jeopardizes the Confidentiality, Integrity, and Availability (CIA triad) of critical data. The rapid evolution of cyber threats necessitates advanced security mechanisms capable of proactive detection and rapid response. Traditional SOCs, primarily reactive and dependent on manual processes, struggle with high volumes of alerts and sophisticated attack vectors. This research paper presents a cost-effective, open-source SOC using Elastic Stack, The Hive, and Cortex for threat detection. The obtained results reduce alert fatigue, cut response times by 40 % and eliminate licensing costs while covering MITRE ATT&CK via Sigma rules. The average obtained is 60 % faster detection versus legacy SIEMs, proving enterprise security can be scalable and vendor-agnostic.
Modeling electromagnetic (EM) wave propagation is critical for achieving efficient and reliable transmissions in maritime communication systems. Indeed, the signal propagation in this environment is strongly affected by several factors including surface roughness, diffraction, and atmospheric ducting. Even though traditional methods like ray tracing and mode theory analysis are computationally efficient, they often fail to capture the irregularities caused by these factors. The parabolic equation (PE) model has proved successful in addressing this issue but remains computationally intensive. This paper proposes a physics-informed neural network (PINN) approach that integrates the PE model with real-world data. The idea is to decompose the complex wave field into real and imaginary components, considering impedance boundary conditions, rough sea reflections, and wide-angle Padé approximation. Simulation results show that the proposed method significantly improves both accuracy and generalizability compared to the conventional models.
