Software vulnerability analysis is critical for maintaining secure and reliable systems, yet traditional Deep Learning (DL) models often act as “black boxes,” lacking transparency and failing to leverage the explicit structural semantics of code. In this paper, we propose KG-HiAttention, a novel neuro-symbolic framework that synergizes sub-symbolic deep learning with symbolic AI-based Knowledge Graphs (KGs). We construct a CPG-inspired lightweight program graph for each software function, approximating control-flow (CFG) and data-flow (DFG) dependencies through line-level edges. This symbolic structure is processed by a Graph Attention Network (GAT) and fused with semantic embeddings from a pre-trained CodeT5 encoder through multimodal fusion (concatenation and MLP classifier). Experiments on the real-world BigVul dataset show that KG-HiAttention achieves competitive performance (AUC-ROC 0.763 ± 0.009, five seeds), statistically equivalent to a strong Hybrid Ensemble baseline, while improving specificity from 0.321 (baseline) to 0.458 and providing graph-based explainability that the baseline cannot offer.

The increasing urgency of carbon footprint reduction in supply chain operations demands innovative optimization approaches that balance economic efficiency with environmental sustainability. This paper presents a novel carbon-aware route optimization framework that integrates machine learning-based emission prediction with genetic algorithm optimization for sustainable supply chain management. Our hybrid approach combines Random Forest and XGBoost models in an optimized ensemble to predict carbon emissions with high accuracy (MAPE: 9.48%, R2: 0.928), while a genetic algorithm optimizes routes considering both cost and carbon constraints. The framework is validated through two complementary scenarios: (1) controlled experiments on synthetic datasets (n=3,500 routes across three network sizes: 500, 1000, and 2000 routes) derived from real-world emission factors demonstrate 19.5% average emission reduction with 4.7% cost increase, and (2) a quasi-real case study on Salamanca regional distribution network (n=12 routes, 776.6 tons CO2e annually) achieves a 41.4% emission reduction with 8.6% cost increase through strategic modal shifts to rail transport. Both scenarios significantly outperform traditional cost-only optimization methods. The proposed approach provides supply chain managers with actionable insights for achieving sustainability goals while maintaining operational efficiency.

Brain tumors pose a significant challenge in medical diagnosis and treatment, requiring precise and efficient techniques for detection, segmentation, and classification. Recent advancements in artificial intelligence (AI), particularly machine learning and deep learning, have revolutionized neuro-oncology by enabling automated and accurate tumor analysis through medical imaging. This study explores the role of AI-driven approaches in brain tumor segmentation and classification. Various AI methodologies, including convolutional neural networks and transfer learning, have demonstrated superior performance in identifying tumor regions, estimating malignancy, and enhancing diagnostic precision. However, addressing some problems of these techniques, such as handling data imbalance and model interpretability, is crucial for integrating AI-based models into clinical practice. This paper provides an in-depth analysis of current AI techniques, performance evaluation metrics, and available datasets, offering insights into future directions for improving AI applications in brain tumor research.

Current Network Intrusion Detection Systems (NIDS) often fail to detect adversarial evasion attacks, creating critical security blind spots. To address this, we propose a standardized adversarial evaluation protocol that quantifies performance degradation against Fast Gradient Sign Method (FGSM), Projected Gradient Descent (PGD), and AutoAttack ensemble attacks, establishing empirically observed performance bounds. We implemented a high-throughput hybrid architecture combining 1D-CNN, Bidirectional LSTM, and Transformer mechanisms, designed specifically to balance varying traffic dynamics and robustness. Unlike prior studies that report only clean-data accuracy, our evaluation of UNSW-NB15, CICIDS2017, and CICIoT2023 demonstrates competitive performance (e.g., strong multi-class F1 scores) while revealing robustness profiles up to an operational limit of epsilon = 0.05. Crucially, we validated our results under a temporal split using the official UNSW-NB15 train/test partition, confirming that binary detection (94.20% accuracy, 95.69% F1) generalizes under distribution shift. We further compared the proposed method with PGD-based adversarial training (PGD-AT) to quantify the robustness-accuracy trade-off. Our results advocate the use of security curves as a standard metric for NIDS validation in hostile environments.

IntroductionThis study examines the role of generative artificial intelligence (GenAI) in cybersecurity threat detection, focusing on its usefulness in workflows that support human decision-making.MethodsExperiments were performed on the BODMAS dataset (134,435 samples) and a smaller exploratory subset of UNSW-NB15. State-of-the-art machine learning (ML) classifiers were compared with a zero-shot large language model (LLM) using standard classification metrics, while also considering latency, cost, and hallucination risk.ResultsML classifiers consistently outperformed the LLM-based system on standard detection metrics. However, the LLM showed value in cases of ambiguity, where it could provide short plain-language explanations, organize alert-related context, and generate initial interpretations for instances that did not match learned classes.DiscussionGenAI is unlikely to replace ML-based detection methods, but it can provide useful interpretive support for ambiguous or unfamiliar alerts. A hybrid pipeline is therefore proposed, in which ML handles high-confidence and time-sensitive decisions, while the LLM is used selectively for low-confidence cases or when explanatory support is needed. Human oversight remains necessary to address hallucination risk and ensure reliability.

This article discusses the proposed study for the completion of a Ph.D. thesis. A system to remotely monitor patient progress in rehabilitation using IoT devices and mobile platforms is proposed. This system feeds advanced artificial intelligence (AI) algorithms and techniques such as federated learning. Additionally, real-time posture estimation with computer vision is implemented. This set of technologies enables accurate monitoring of rehabilitation exercises without the need for constant supervision. This approach is intended to benefit both the healthcare professional, by reducing the burden associated with routine check-ups, and the patient, who can be assured of continuous monitoring of their injury and a rapid response to any abnormalities in recovery.

This study presents the design and implementation of a single-institution intelligent diagnostic system to identify low mid-period academic performance, aimed at activating proactive and preventive tutoring before a final assessment. The system features an integrated analytical architecture comprising an inferential framework, a predictive framework, an explainability framework, a validation framework, and a Streamlit-based web prototype. The sample uses 18,604 longitudinal academic records from 1264 unique students enrolled across 7 consecutive academic periods (2017-2020) at an Ecuadorian university. Results indicate that curricular level is the structural predictor with the greatest independent contribution (semi-partial R2 = 0.044), followed by academic period (semi-partial R2 = 0.026). Random Forest achieved the best overall performance (MAE = 1.267 +/- 0.04; RMSE = 1.714 +/- 0.05; R2 = 0.551 +/- 0.02), outperforming other algorithms. SHAP explainability confirms the primacy of curricular level and academic period as individual-level risk-associated factors, enabling the generation of interpretable alerts for tutors. The equity analysis revealed that students aged 30-50 years (ratio = 1.375) and the province with code 18 (ratio = 1.395) constitute priority subgroups for data enrichment prior to institutional deployment. External validation with real users is identified as the next research stage.

Reinforcement learning (RL) has emerged as a pivotal technology for autonomous driving, providing a framework where an agent learns optimal decision-making through environment interaction without pre-defined labels. This review focuses on the application of RL in autonomous driving systems, exploring key concepts such as value-based methods, policy-based methods, and actor-critic methods within the context of Markov decision processes (MDPs) and partially observable Markov decision processes (POMDPs). We discuss the integration of these methods in navigating the complex, dynamic environments encountered in autonomous driving, addressing challenges in perception, sensing, decision-making, and control. The review further evaluates the role of simulators in testing and refining RL algorithms, highlighting their importance in the development cycle of autonomous vehicles.

Convolutional Neural Networks (CNNs) have proven effective in automated plant disease diagnosis, significantly contributing to crop health monitoring. However, their limited interpretability hinders practical deployment in real-world agricultural settings. To address this, we explore advanced CNN architectures, namely ResNet-50 and EfficientNet, augmented with attention mechanisms. These models enhance accuracy by optimizing depth, width, and resolution, while attention layers improve transparency by focusing on disease-relevant regions. Experiments using the PlantVillage dataset show that basic CNNs achieve 46.69% accuracy, while ResNet-50 and EfficientNet attain 63.79% and 98.27%, respectively. On a 39-class extended dataset, our proposed EfficientNet-B0 with attention (EfficientNetB0-Attn), integrating an attention module at layer 262, achieves 99.39% accuracy. This approach significantly enhances interpretability without compromising performance. The attention module generates weights via backpropagation, allowing the model to emphasize disease-relevant image regions, thereby enhancing both accuracy and interpretability.

Multiple open data portals offer data that may not appear to violate data privacy or confidentiality laws at first glance. However, a thorough study of these datasets and their relationships with others reveals that confidential or private information may be obtained in certain cases. To address these issues, this article proposes a solution that involves implementing a series of AI-powered modules. The goal of these modules is to analyze the quality of the data and its potential combinations with linked data that could lead to legal non-compliance or data quality issues. Due to the lack of standardization across different open data portals, this model facilitates the improvement of these portals for information extraction and decision-making purposes while ensuring compliance with data privacy and confidentiality laws.

We present a novel architectural framework for oracles in Distributed Ledger Technology (DLT), crafted to enhance the interaction between smart contracts and external data sources. Oracles play a crucial role in bridging the deterministic nature of blockchains and the dynamic external environment, enabling smart contracts to access real-world data required for their execution. Our architecture introduces a modular design, consolidating data storage and processing to enhance scalability and security throughout the system. The architecture facilitates autonomous task execution crucial for real-time smart contract functionality, activated by customizable conditions to accommodate various use cases ranging from decentralized finance to supply chain management. To promote decentralization and enhance fault tolerance, we recommend establishing oracle networks that utilize robust consensus mechanisms for data consistency and security. Security measures embedded in our framework involve the incorporation of advanced encryption, isolated operational modules, and the utilization of private nodes to safeguard against data tampering and unauthorized access. These tactics safeguard the integrity of data transactions and the dependability of the oracle network. Our proposal aims to increase the significance of oracles within DLT settings, with the goal of promoting broader acceptance and effectiveness in operations. The proposed architectural approach is positioned to offer a solid positioning for the progression of decentralized applications, guaranteeing the safe and dependable incorporation of external data into blockchain ecosystems.

Human interaction has changed considerably with the emergence of the Internet. Today, a large percentage of daily communication takes place on instant messaging applications and social networks. In fact, there has been a considerable increase in the use of social networks because new social networks are being created for specific purposes, such as the search for employment or tourism. When the amount of content on a social network is large, it is necessary to help users find content of their interest. In this regard, artificial intelligence techniques can greatly facilitate the task of searching for relevant content. This paper presents a recommender system for a business and employment oriented social network, on which users are recommended job offers and other user profiles to follow. The presented system is based on virtual organizations of agents, and uses artificial neural networks to determine whether job offers and users should be recommended or not. The system has been evaluated on a real social network; its recommendations regarding job offers and user profiles have had a high acceptance rate.

The use of technologies like artificial intelligence can drive productivity growth, efficiency and innovation. The goal of this study is to develop an anomaly detection method for locating flaws on the surface of sandwich panels using YOLOv5. The proposed algorithm extracts information locally from an image through a prediction system that creates bounding boxes and determines whether the sandwich panel surface contains flaws. It attempts to reject or accept a product based on quality levels specified in the standard. To evaluate the proposed method, a comparison was made with a sandwich panel damage detection method based on a convolutional neural network and methods based on thresholding. The findings show that the proposed method, which is based on an object detector, is more accurate than the alternatives. The characteristics of the model, which can reject or accept a product according to the standard and limit allowable manufacturing flaws to obtain a quality product, also enable this system to improve industrial standards for producing sandwich panels while increasing speed.

Recommendation systems were implemented as a solution to reducing the time and effort required by a user to search for information. In the development of the recommendation systems that offer the best performance, artificial intelligence algorithms are used, in combination with various recommendation approaches. However, in the educational context these systems have a different connotation

There may be multiple reasons for tooth extraction, such as deep cavities, an infection that has destroyed an important portion of the tooth or the bone that surrounds it, or for orthodontic reasons, such as the lack of space for all the teeth in the mouth. In the case of orthodontics, however, there is a relationship between tooth extraction and the craniofacial morphological pattern. The purpose of this study is to establish whether such a relationship exists in adolescents and to evaluate it and to serve as a tool to support medical decision making. Machine Learning techniques can now be applied to datasets to discover relationships between different variables. Thus, this study involves the application of a series of Machine Learning techniques to a dataset containing information on orthodontic tooth extraction in adolescents. It has been discovered that by following simple rules it is possible to identify the need of treatment in 98.7