Earticle

Data modeling in data warehousing is a multifaceted process crucial for effective data management and analysis. Through techniques like conceptual, logical, and physical modeling, data architects organize complex data structures. The iterative nature of modeling allows adaptation to evolving business needs. The analysis shows that most organizations used the hybrid approach to create the data modeling in the data warehouse because this approach helps to create a good and efficient model of data in the data warehouse. But on the other hand, when we talk about the anomalies, researchers used an isolation forest and an LSTM algorithm based on net earnings by month. This is the way that helps to remove maximum anomalies from the data warehouse during storing data from sources. In this era, mostly people use Big data concept for data model for data warehousing but this technique is much complex for storing data and retrieval of data in this technique just single thing is missed for the data modeling named as reusability in future working of big data if we majorly emphasize on reusability of data the this tech can be most efficient as compare to present.

Advances in artificial intelligence (AI) have been instrumental in developing smart homes, increasing efficiency, effectiveness and user experience. Artificial Intelligence algorithms help smart homes understand user behavior, preferences and routines. Thanks to continuous learning, these systems can adapt to changing environments and perform tasks such as adjusting lighting, temperature and security according to individual behavior. AI in smart homes uses analytics to predict customer needs. By analyzing historical usage patterns, weather, and other factors, AI can predict customer preferences, improve resource utilization, and create better living conditions. Integration of NLP allows users to interact with smart home devices using natural language. This increases accessibility and facilitates user engagement, as the AI-powered voice assistant can understand and respond to commands, making management of various devices easier. Artificial intelligence algorithms help improve energy management in smart homes. Artificial intelligence systems can recommend and implement energy-saving measures by analyzing energy consumption patterns, thus reducing energy costs and environmental footprint. Artificial intelligence-supported security systems provide smarter threat detection and response. Machine learning algorithms improve the accuracy of home security assessments by distinguishing between normal and suspicious activity. Facial recognition and behavioral analysis provide better access control. Artificial intelligence supports the interaction of different smart devices and platforms, creating an integrated and connected system. This interaction allows different devices to integrate and share information, leading to the overall performance of the smart home. User personalization: Artificial intelligence tailors smart home experiences to personal preferences. From adjusting lights and music to predicting when home appliances will turn on, AIpowered personalization increases user comfort and satisfaction, creating an experience that is immersive, great to be around, and fun. The integration of artificial intelligence and smart home is leading the revolution in home automation.

This paper proposes a vision-language model for the joint recognition of Low Probability of Intercept (LPI) radar signals through time-frequency distribution (TFD)-text alignment. The proposed framework unifies waveform classification and signal parameter estimation by aligning TFD spectrograms with hierarchical textual prompts in a shared embedding space. To support both general waveform type recognition and fine-grained parameter inference, we introduce a prompt dropout strategy that balances rich and simple prompts during training. Evaluated on multiple TFD representations including SPWVD, CWD, and SAFI, the model demonstrates high accuracy and interpretability across both tasks. This unified approach offers a compact, extensible solution for LPI radar signal understanding.

Small Language Models are competitive without large-scale infrastructure, their performance is highly contingent on prompt design. This study analyzes the sensitivity of BitNet b1.58-2B-4T to label exposure and fewshot exemplar composition on a 36-class medical query classification task. We generated 504 items consisting of 6 direct and 8 indirect questions for each disease and after removing cross-exemplar leakage the final evaluation set contained 494 items. With no parameter updates, 0/1/2/5/10- shot prompting was evaluated using Accuracy. Under the nolabel- exposure setting accuracy increased as more exemplars were provided. However, these gains were accompanied by growing prediction concentration on exemplar labels. In contrast with label-exposure, zero-shot achieved the highest accuracy, while the inclusion of exemplars reduced accuracy and amplified label bias. These results show that the structure of the prompt tends to shift few-shot effects from beneficial to detrimental. This highlights the importance of controlled prompt design and domain-adaptive training to ensure trustworthy performance.

The efficient deployment of robot-based manufacturing systems is frequently hindered by the substantial time required for programming collision-free robot paths during the commissioning process. This challenge involves intensive tasks such as teach-in, offline programming, and subsequent path optimization. To dramatically accelerate this critical stage, the industry needs an automatic and intelligent path planning system. This work introduces a novel system designed for the autonomous path planning of industrial robots. We conduct an explicit comparison between samplingbased methods such as probabilistic roadmaps (PRM) and rapidly exploring random Trees (RRT), and computational intelligence (CI) based methods, particularly genetic algorithms. Our findings demonstrate the potential for these advanced techniques to drastically reduce robot deployment time.

In the metaverse-- the threat verse, the user identity security is not suffice. The user identities linked to the traversing avatars utilize biometric authentication to authenticate and dictate their actions in the metaverse. Biometric authentication demands input in the form of facial, voice, iris/gaze, and physiological behavioural patterns for its user. Additionally, combining these biometrics with neurometrics for enhanced authentication is being explored. These Neurometric-based authentication systems are less discovered due to their complexity and practicality. However, these systems are stabilized by consuming the Artificial Intelligence (AI) Model Fusion. These systems are unprobed towards their sustainability, security, and usability. Therefore, we attempted to explore the Neurometric-based authentication systems stabilized with AI model fusions. We acutely examined these systems’ infrastructure and its susceptibility to existing threats. This led us to introduce the absent components to be included in the system infrastructure to increase its potential towards probable threats. However, we conclude our study exploring this new direction of possibilities for a Neurometricbased authentication system for the virtual world environment.

This study proposes a multimodal hybrid framework that integrates corporate disclosure texts, stock prices, and market indicators for enhanced stock price prediction and portfolio optimization. Long Short-Term Memory (LSTM) networks are employed to capture temporal dependencies from stock and market data, while FinBERT is used to extract semantic embeddings from disclosure documents. These heterogeneous modalities are fused within a classification model to evaluate the significance of corporate and market events on stock movements. Experimental results reveal substantial performance variation across industries, motivating industry-specific investment strategies. Portfolio optimization based on these multimodal strategies achieves superior riskadjusted returns compared to benchmark approaches.

Accurate yet accessible prediction of Alzheimer’s disease (AD) remains difficult because advanced imaging and cerebrospinal fluid assays are expensive and invasive. This study explores whether two routinely collected clinical measures— patient age and Mini-Mental State Examination (MMSE) score—can form a reliable, interpretable baseline for AD risk modeling. A subject-level cohort of 4,651 participants from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) was analyzed using a logistic-regression framework with z-score scaling and balanced class weighting. Five-fold stratified crossvalidation produced stable performance (AUC = 0.929 ± 0.007, precision = 0.699 ± 0.007, recall = 0.844 ± 0.016, F1 = 0.765 ± 0.004, accuracy = 0.86 ± 0.004). Despite using only two features, the model approached the performance of several reported multimodal systems in discriminative power while remaining transparent and reproducible. This minimal benchmark establishes a reference point for future multimodal extensions and demonstrates that cognitively based screening can achieve clinically meaningful accuracy in resource-limited settings.

This study compares two embedding-based natural language processing techniques—Sentence-BERT (SBERT) combined with HDBSCAN clustering and BERTopic modeling—for detecting complex emotions in short Korean online comments. Using 33,531 comments collected from a YouTube relationship counseling channel, we examined how each method captures nuanced and overlapping sentiments such as affection, avoidance, and conflict. Both models used identical SBERT embeddings and UMAP-based dimensionality reduction, and their clustering performance was quantitatively evaluated using Silhouette Score, Davies–Bouldin Index (DBI), and Calinski–Harabasz Index (CHI). The results show that BERTopic achieved higher coherence and clearer topic boundaries (Silhouette = 0.40, DBI = 0.85, CHI = 15,157) compared to SBERT–HDBSCAN (Silhouette = –0.23, DBI = 1.49, CHI = 1,230). Although both methods yielded high noise ratios due to the leaf-based density clustering, BERTopic effectively reclassified semantically relevant comments through its ClassTF-IDF weighting, improving topic stability and interpretability. These findings suggest that BERTopic provides superior performance for analyzing short, emotion-rich Korean text and offers methodological insight for future sentiment analysis research. This electronic document is a “live” template and already defines the components of your paper [title, text, heads, etc.] in its style sheet.

Early diagnosis of dementia, such as Alzheimer's disease, is clinically crucial, and deep learning-based brain MRI analysis has gained significant attention. However, acquiring large-scale datasets remains challenging in medical imaging, necessitating learning from small datasets. This study quantitatively compares the dementia classification performance of CNN-based ResNet-18 and various Transformer architectures (Swin Transformer, Vision Transformer, UNETR) using the OASIS-3 brain MRI dataset (3,428 samples) and systematically evaluates their suitability for small-scale medical data. Methods: Data comprising 2,943 normal (85.8%) and 485 dementia (14.2%) cases were split 80:20, with 5-fold cross-validation. Focal Loss was applied to mitigate class imbalance, and metrics including Sensitivity, Specificity, Balanced Accuracy, and AUC-ROC were evaluated. Results: ResNet-18 achieved the most balanced performance with Sensitivity of 79.38%, Balanced Accuracy of 77.38%, and AUC-ROC of 85.40%. Transformer models showed distinctly different patterns: Swin Transformer (Sensitivity 42.27%, AUC 81.78%) exhibited normal-class bias, Vision Transformer (Sensitivity 22.68%, AUC 47.89%) nearly failed to learn with pure global attention, and UNETR (Sensitivity 88.66%, AUC 73.00%) achieved highest sensitivity but severely low specificity (41.77%). ResNet-18 also demonstrated superior learning efficiency (parameters: 33M, training time: 2.1h) compared to Transformers (Swin: 60-70M/3.1h, ViT: 18M/3.0h, UNETR: 10- 12M/4.0h). This study confirms that CNN's inductive bias and structural efficiency are more effective than Transformer's global attention or hybrid approaches in small-scale brain MRI datasets, with ResNet-18 proving most suitable for dementia screening due to balanced sensitivity-specificity trade-off.

This study proposes an early-warning framework for attrition risk that combines unsupervised clustering (K-means, Gaussian Mixture Model) with survival analysis (Kaplan–Meier, Cox proportional hazards), using data on 11,090 wage workers from the Korean Labor and Income Panel Study (KLIPS, 2002– 2021). After segmenting latent heterogeneity with K-means and GMM, we estimated cluster-specific survival functions and risk factors via KM survival curves and Cox regression. Based on internal and model-based criteria (silhouette, Calinski–Harabasz, BIC, ARI), performance was optimal when the number of clusters was set to five, with a silhouette coefficient of 0.247, a Calinski– Harabasz index of 680.3, the lowest BIC, and an ARI of 0.94. The 24-month cumulative attrition probability for Cluster 2 (low satisfaction and overwork) was approximately 50 percent (highest), whereas Cluster 1 (high satisfaction and stable) was approximately 12 percent (lowest), with a significant difference by the log-rank test, with the p-value below 0.001. In the Cox analysis, Cluster 2 showed a hazard ratio (HR) of about 2.0, and declines in job satisfaction and increases in overtime hours emerged as significant risk factors. The proposed framework provides quantitative grounds for targeted interventions in high-risk segments by means of early-warning indicators based on 12- and 24-month attrition probabilities and lift.

This study presents a deep-learning-based framework integrating a single heat-flux sensor for rapid and accurate estimation of core body temperature (CBT). A gated recurrent unit (GRU) model was trained using a large-scale dataset comprising finite-element simulations and experimental transient responses to learn conduction- and convection-governed features. By analyzing only the initial 30 s of temperature signals, the model achieved a mean absolute error below 0.1 °C across a wide range of ambient temperatures (5–35 °C), convective heat transfer coefficients (0–50 W·m⁻²·K⁻¹), and skin conductivities (0.32–0.50 W·m⁻¹·K⁻¹). This data-driven approach eliminated the need for prolonged thermal stabilization, enabling site-independent CBT prediction with reliable performance under dynamic environmental and physiological conditions.

Out-of-Stock (OOS) detection is a critical task in retail management, yet traditional automated methods suffer from significant limitations. Existing approaches, whether based on classical image processing or closed-set deep learning models, lack scalability and robustness, often requiring extensive retraining for every new shelf layout or product type. This paper proposes a novel and flexible OOS detection pipeline that leverages the power of Open- Vocabulary Object Detection (OVD). Instead of attempting to directly detect "empty space," our method first identifies all present products using an OVD model guided by flexible text prompts. An inverse occupancy mask is then generated to identify potential OOS regions, which are subsequently refined through a robust multi-stage post-processing filter. Experiments on our custom dataset of 149 diverse retail shelf images demonstrate the superiority of this approach. Our method achieves an OOS detection Accuracy of 87.9%, vastly outperforming the baseline approach (70.1% Accuracy). Furthermore, by applying per-shelf optimized prompts and parameters, our model's Accuracy increases to 96.84%, highlighting its high adaptability and effectiveness for realworld retail environments.

The evaluation of Digital Elevation Model superresolution (DEM-SR) has predominantly relied on PSNR, often overlooking practical performance in physics-based VR simulations. Rather than proposing a new architecture, this work presents a quantitative case study evaluating existing Vision Transformers (SwinIR and Swin2SR) on real-world Copernicus GLO-30 terrain data. We assess their performance using both PSNR and application-centric metrics that capture statistical consistency and in-VR usability. Our analysis reveals that despite achieving respectable PSNR scores, architecturally advanced models can exhibit pronounced instability and immersion-breaking artifacts. We conclude that PSNR is insufficient in isolation and argue for incorporating practical, application-oriented metrics to ensure genuinely usable VR terrain.

Metaverse, our virtual reality, is traversed via an Avatar linked to a user profile through Personal Identifiable Information (PII). To secure this PII from causing attacker infiltration, only authorised users should access these avatars permitted by the authentication systems. The authentication systems are researched to be resilient against attackers’ manipulations. These systems rely on dynamic and real-time sensor data rather than static information from the user for authentication. Dynamic sensor data captured through Head Mounted Displays (HMDs) is highly classifiable with Machine learning (ML) and Deep Learning (DL) algorithms. Over time, the model training requires an upgrade through evolution in data processing and learning. Self-learning —Adaptive learning can lead this system to transform with its learn-evolve-adapt learning strategy. Therefore, our study attempts to explore authentication systems developed for HMDs, capturing realtime dynamic sensor data. With its results, we concluded that these systems are highly sensitive while processing the sensor data. We list out the risk factors of utilising adaptive learning for an authentication system based on neurometric data combined with biometric data. This study will be the state of the art for the self-learning algorithms for biometric and neurometric data-based authentication systems.

In this study we evaluate a production-style OpenTelemetry (OTel) pipeline on Google Kubernetes Engine (GKE) Autopilot under a sustained trace workload, instrumenting end-to-end ingestion→processing→export and scraping Collector self-telemetry and spanmetrics with Prometheus over a 30-minute run. The Collector averaged 573 spans/s accepted and 561 spans/s exported, yielding 97.84% within-window export efficiency and peaking near 924 spans/s. A brief saturation interval (150 s) produced a queue peak of 6.38, short drops (peak 3.49 spans/s, ~358 total), and p95 inflation to 116 ms; median latency remained low (p50 ≈ 12.9 ms) and recovered after pressure subsided. Resource footprint was modest (≈0.36 CPU cores, 0.21 GiB memory, sub-Mbps network), indicating headroom. We document integration pitfalls (OTLP endpoint/protocol) and show that queue growth and exporter errors anticipate p95 tails. The study contributes reproducible methodology, quantitative evidence of cloud-scale OTel scalability, and operator guidance for capacity planning and alert design.

This paper presents an AI-empowered fluid antenna system (FAS) framework to enhance the robustness of positioning, navigation, and timing (PNT) in future sixthgeneration (6G) satellite terrestrial networks. The FAS employs movable or reconfigurable liquid-metal antennas that dynamically alter geometry and port position to exploit spatial diversity and mitigate multipath fading. Building on verified FAS fundamentals, exposure-constrained optimization, and a fluid-antenna multiple access (FAMA) model, a lightweight AI controller is proposed for real-time port selection using carrierto- noise ratio, signal-to-noise ratio, and Doppler-shift features. The proposed framework integrates safety constraints and enables adaptive port hopping for seamless hybrid global navigation satellite system (GNSS) and 6G PNT operation. Recent research on movable-antenna integrated sensing and communication (ISAC) and received-signal-strength-indicatorbased (RSSI-based) FAS positioning demonstrates the feasibility of AI-driven FAS for high-precision timing and localization in next-generation networks.

Email spoofing represents a common cybersecurity risk that abuses the weaknesses in the email protocols to falsify the sender addresses and trick the recipients into providing sensitive information or performing malicious requests. Conventional rulebased detection systems have been found ineffective against more advanced forms of spoofing. This research project suggests using machine learning to identify email spoofing using a range of features derived through email header, content, and metadata. Various algorithms, such as Support Vector Machines (SVM), Random Forest, and Logistic Regression, are tested to find the most suitable in terms of identifying legitimate emails and spoofed ones. On the dataset, preprocessing is done through tokenization, feature encoding, and vectorization to increase the model accuracy. The evaluation of the performance is performed in terms of precision, recall, F1-score, or ROC-AUC. Through experiments, it has been shown that machine learning models, especially ensemble based methods, greatly exceed traditional methods in accuracy and low false positive rate in detecting spoofed emails. This work has demonstrated the promise of intelligent systems when used to reinforce email security and has also offered a scalable implementation of real-time detection of spoofing.

The main visual feature is surface cracks, which are caused by loopholes that are embedded into the structures due to manufacturing faults and overloading factors. The structural health monitoring must be precise and more efficient for detecting surface cracks in concrete structures. Human inspection is used to identify damage on concrete surfaces. But, these traditional visual observation techniques are not more effective for large concrete structures. Moreover, this outmoded human labor practice for crack detection is intensive, expensive, and inefficient. To predict potentially hazardous situations caused by cracks on concrete surfaces, it is crucial to have an efficient, fast, and well-organized inspection system for concrete surface cracks. The automatic crack detection system must be effective in identifying cracks, damage, and segmenting them. In this research, a deep learning (DL) algorithm model is active for crack detection in concrete structure images to assess the influence on structural health. This design work is projected to provide a fast and active solution for identifying dust/duct type to prevent power losses using an image classification model based on DL. The VGG-16 DL model significantly analyzes the precision and accuracy of identifying the crack surfaces on concrete structures. The exceptional performance of the projected model achieves a training accuracy of 99.99% and a test accuracy of 99.96%, with an F1 score of 0.9995, precision of 0.9995, and a sensitivity of 0.9997. This is more precise, costeffective, and more efficient than human resources to find the defect on concrete surfaces that may support the healthy life of concrete structures.

Modern warfare demands tactical networks that can survive rapid Electronic Warfare adaptation. This study proposes a 5G Femtocell Drone Interconnected architecture utilizing Integrated Access and Backhaul and Network Slicing to establish self healing connectivity without wired infrastructure. The system addresses scalability and precision challenges through a Group Handover scheme and an adaptive navigation fusion of PTPsec and Visual Inertial Odometry. Furthermore, context aware Adaptive Policies are implemented to balance Zero Trust security with rapid network recovery. Simulation results confirm that this approach reduces signaling overhead by 86 percent and shortens link recovery time by 57 percent, effectively satisfying tactical requirements for survivability and cognitive superiority in GPS denied environments.

We present an unsupervised anomaly detection framework for aluminum molten metal processes that trains exclusively on normal operational data. The extreme temperatures and dynamic luminance variations characteristic of molten metal environments make it practically infeasible to collect representative samples of anomalies—such as foreign material contamination, thermal irregularities, or flow inconsistencies—and manual annotation remains prohibitively labor-intensive. Our approach employs a Convolutional Autoencoder to capture the visual and thermal signatures of normal process states. Anomalies are detected by computing reconstruction errors and comparing them against thresholds derived from the upper percentiles of the normal error distribution. Experimental validation across multiple production lines demonstrates that our method achieves high detection accuracy even with limited abnormal samples, offering a practical solution for automated quality control and real-time process monitoring in industrial casting operations.

This paper proposes a practical, privacypreserving healthcare data-sharing architecture that integrates a local CDM(Common Data Model) with a FHIR (Fast Healthcare Interoperability Resources)–based self-serve interface under a federated data-mesh governance model. Each organization operates autonomously as a domain team and maintains its own data lake. In the upload flow, source data are standardized into a local CDM and passed through a De-ID gate for pseudonymization, date shifting, generalization, and freetext PII redaction before being stored in a Processed-DeID zone. In the download flow, only cataloged and validated data products are exposed as FHIR resources, where context- and authorization-aware dynamic masking is applied, with support for Bulk Export and incremental queries. Federated computational governance enforces standards, security, and change management as policy-as-code, ensuring consistent compliance at deploy time and runtime. Internally, CDM ensures data quality and consistency; across organizations, FHIR simplifies data contracts and standardizes discovery, authorization, and use. The architecture enables safe exchange and reuse even among institutions with different CDMs, while improving lineage visibility, accountability, operational scalability, and change predictability. This two-layer model can serve as a reference architecture for accelerating data flows across clinical, research, and industry collaborations.

In this study, we propose a semi-supervised dataset curation framework that leverages both high-confidence labeled protein sequence data and automated weakly labeled protein sequence data to refine dataset quality prior to model training. The approach centers on using a pre-trained ProtBERT model to iteratively assign pseudo-labels to uncertain samples, followed by subsequent model retraining, with the goal of enhancing robustness and generalization. We anticipate that a curated dataset constructed in this way will significantly enhance toxin-classification performance— measured in accuracy, F1-score, and MCC—compared to models trained solely on manually labeled or automatically annotated data.

The merger of biotech and cyberwarfare has opened up a new war zone where DNA is programmable, weaponizable, and exploitable for gaining the upper hand in war. The paper delves into the interdisciplinary field of cyberbiosecurity which is an integrated framework that combines cybersecurity, synthetic biology, and defense strategy to deal with biological cyber threats. Once cyberbiosecurity is breached, and AI-driven genome editing is used along with a clandestine bioinformatics operation, an individual, a population, or an ecosystem can be the target with surgical precision. The study recognizes the means for genetic weaponization, pinpoints the genomic databases' weaknesses, and suggests a bio-digital defense model that combines secure DNA sequencing, blockchainbased genome storage, and AI threat detection to overcome the challenges.This paper, by defining DNA hacking as a component of national defense strategy, is an effort to equip the military and scientific communities with the requisite knowledge and skills to combat the rise of genetic warfare that is bound to happen.

The research is about an AI-powered egovernance framework that aims at transforming the national defense and security sector with the help of digital intelligence. The restructured system features artificial intelligence, data analytics, and secure governance layers to facilitate decision-making, threat detection, and resource management. From a methodological point of view, the paper represents risk as influenced by cognitive bias and control effectiveness, based on hologram visuals and key performance indicator (KPI) analyses. The report illustrates that the adoption of AI leads to the achievement of tasks with much higher accuracy, reduction of the time interval between two events, or steps in a process and making it easier for audit trail to be followed in defense operations. The findings support the view that the combination of bias reduction and control effectiveness improvement leads to a decrease in operational risk, thereby paving the way for digital transformation in national defense governance to be conducted in a secure and a way that can be easily understood by human users.

The use of social engineering has become the primary cause of security breaches in the digital world, where it takes advantage of human nature instead of the weaknesses in technology. Social engineering, unlike a normal hacker, uses charm, lies, and psychological stress on the victim to get him/her to reveal confidential information. This research uncovers deceiving tactics used in social engineering to examine the main attack vectors like phishing, pretexting, quid pro quo, baiting, and tailgating, and to show the effect of these tactics on companies. It cites examples of the past to convince the reader to resort to the solution of awareness programs, well-structured security policies, and staff training to counter the threat.

The dramatic increase in cyberattacks targeting government agencies, critical infrastructure, and defense contractors has revealed the inefficiencies of traditional software development life-cycle (SDLC) models. Most SDLC models do not indicate security explicitly, hence secure software development practices need to be infused into each phase so as to limit the exploitation of security loopholes and safeguard the security of high‑security systems. This document puts forward SSDF‑HS, a secure software development framework for high‑security organizations. The comparative study of various hypothetical projects reveals that SSDF‑HS significantly decreases vulnerability density and mean time to remediate while also enhancing compliance scores. The mathematical risk‑reduction model illustrates the lowering of residual risk when security is integrated early (shift‑left). The framework assists high‑security organizations in writing software that is inherently secure and can withstand future threats.

Decentralized apps are now much more accessible thanks to AI-powered in-app support. Submenu after submenu is not required, nor is it necessary to guess what to click next. Come on, though, these helpful AI features bring with them some new problems. Code injection, data poisoning, and tricks that change what you see or click can all be successful. Prompt injection is a technique that attackers can use to give the AI malicious commands directly. In fact, this study explores the mechanisms behind these attacks. It looks at real-world situations, talks to security professionals, and sorts through the results of real security audits. Most problems show up directly in front of the user interface, whether it's malicious scripts or phony prompts.

Bringing together artificial intelligence (AI) and blockchain technology in the NFT space has not only changed the overall idea of digital ownership but also opened up new avenues for content authentication, asset management, and smart contract-based management through automation. The expansion of this technology, however, also brought security issues to the fore. Among these concerns are side-channel leakage, identity spoofing, and man-in-the-middle (MitM) attacks. This work proposes a detailed AI-supported NFT security framework that brings together three main technologies: X.509 Certificate Validation, CacheShield, and Context-Based Node Acceptance (CBNA), forming a system that is secure, private, and trusted. The X.509 system fights against replay and impersonation attacks by authenticating blockchain nodes and users, and also TransCert validation. CacheShield blocks the leakage of confidential information by determining the cache contention using K-Means clustering and hardware performance counters, while the CBNA algorithm (Random Forest Classifier) keeps secure connections in the SDN network by detecting malicious nodes. The outcomes revealed that the system significantly boosts performance in terms of accuracy, authentication latency, and tolerance of attacks. This framework not only fortifies the trust and security of decentralized AI-NFT applications but also opens the door to quantum-safe and context-aware blockchain systems in the future.

Management of Population movement among cities is important for many sectors, such as urban planning, emergency management, and traffic management, etc. The early approaches, including radiation and gravity, give the primary insights but struggle to capture the complex topological and directional nature of inter-city mobility networks. This paper presents a Graph Convolutional Network (GCN)-based framework for modeling and predicting population movement between Chinese cities using Baidu mobility data. Cities are represented as nodes in a directed graph, with weighted edges indicating monthly outbound flows. A multi-layer GCN learns node embeddings that encode both local and global spatial dependencies, enabling the prediction of continuous relationship scores that reflect the intensity of movement. Experimental results, MedAPE of 5.59 and MAPE of 19.36, as well as relationship scores from major cities such as Shanghai and Shijiazhuang, demonstrate that the model effectively identifies stable mobility corridors and evolving connections over time. Overall, the proposed approach provides interpretable insights into population mobility dynamics and supports data-driven decision-making in urban forecasting and regional policy design.