Blockchain fraud detection confronts escalating challenges in Ethereum ecosystems where evolving fraud patterns result in billions of dollars in annual losses, yet existing ap proaches fail to adapt without catastrophic forgetting while pro viding no theoretical guarantees for financial system deployment. This paper presents AHGT-DFD, a framework integrating four ML techniques automated pattern discovery through Dirichlet process priors, hierarchical variational learning, heterogeneous graph transformers, and continual learning mechanisms into a theoretically grounded system. We provide rigorous theoret ical foundations including PAC-learning bounds guaranteeing generalization performance within 6.21% of empirical error with 95% confidence, polynomial-time convergence analysis, and certified robustness against adversarial perturbations. Compre hensive evaluation on real-world Ethereum datasets demonstrates 4.62 percentage point F1-score improvement for Ponzi detection (95.58% vs. 92.15% best baseline) and 4.94 percentage point improvement for phishing detection (97.41% vs. 94.47% best baseline). The framework maintains 94.2% performance reten tion on evolving patterns compared to 78.6% for conventional approaches, while supporting real-time processing with sub-15ms inference latency suitable for production blockchain security systems. This integration with theoretical guarantees provides a foundational advance for dependable fraud detection in adver sarial financial environments.

Overtourism creates unsustainable concentration patterns at tourist destinations, degrading visitor experiences and site preservation. This paper presents a multi-agent reinforcement learning framework employing QMIX coordination with curriculum learning to balance individual tourist satisfaction with collective crowd distribution objectives. The system integrates Rainbow DQN agents with transformer-based communication, progressively scaling complexity from 6 to 18 points of interest to facilitate coordination learning. Evaluation on 2.7 million VeronaCard visits demonstrates satisfaction scores of 83.4±1.9 compared to 65.2±3.3 for historical patterns (28% improvement) while reducing crowd inequality (Gini coefficient) from 0.30 to 0.16. The framework achieves 69% point-of-interest coverage and 68.4% overcrowding reduction. Training requires 230 GPU-hours as a one-time investment, while inference operates at 47ms latency, enabling real-time deployment. Statistical validation through Friedman testing yields p−9. Implementation available at https://anonymous.4open.science/r/QMIX-Based-Multi-Agent--0F47/README.md.

The rapid growth of blockchain networks has introduced unprecedented challenges in detecting anomalous activities within complex transaction graphs. While existing approaches struggle with multi-edge scenarios and temporal dependencies, we present SentinelGNN, a novel graph neural network architecture that achieves state-of-the-art performance in detecting five types of blockchain anomalies: point, contextual, collective, temporal, and structural. Our key technical innovations include: (1) a temporal-aware edge sampling mechanism that effectively handles multiple transaction edges while preserving critical temporal information, (2) a dual-stream architecture that separately processes structural and temporal patterns through specialized neural pathways, and (3) an adaptive gating mechanism that dynamically fuses topological and feature information based on their relative importance. Through extensive experimentation on a large-scale Ethereum dataset containing 6.08M nodes and 38.90M edges, SentinelGNN achieves ROC-AUC scores of 0.980 ± 0.01 and PR-AUC scores of 0.975 ± 0.01, outperforming traditional graph neural networks by significant margins (27% on point anomalies, 25% on contextual anomalies, and 22% on collective anomalies). The model demonstrates particular strength in handling temporal pattern deviations while maintaining computational efficiency, processing 100,000 transactions per second. Comprehensive ablation studies validate the effectiveness of each component, with the temporal-aware sampling providing a 15% improvement in pattern recognition accuracy. Beyond blockchain security, SentinelGNN shows promise in financial fraud detection and supply chain monitoring, offering a scalable foundation for securing decentralized systems.

Non-fungible token (NFT) markets present a dual analytical challenge: integrating heterogeneous data modalities (high-dimensional visual features and discrete transaction sequences) while preserving privacy for sensitive wallet addresses and trading strategies. Current approaches analyze visual attributes or transaction patterns in isolation, missing critical value drivers from cross-modal interactions. Meanwhile, existing multimodal techniques lack formal privacy guarantees, exposing participants to inference attacks. This article introduces PrivaMod, a privacy-preserving Bayesian framework that addresses these limitations through uncertainty-aware multimodal fusion. Our approach implements precision-weighted Bayesian fusion that dynamically adjusts modality contributions based on quantified uncertainty levels, while integrating Rényi Differential Privacy throughout the pipeline via calibrated noise injection and adaptive gradient clipping. Evaluated on 167,492 CryptoPunk transactions, PrivaMod achieves a market efficiency score of 0.874 and R 2 of 0.912, outperforming existing methods by 13.4% through superior cross-modal integration while maintaining strong privacy guarantees ( \(\varepsilon\) = 0.08, \(\delta\) = 1e-5) with membership inference attack success rates near random guessing (53.4%). The system demonstrates that privacy-preserving techniques can enhance rather than compromise analytical performance, establishing a foundation for responsible market analysis. To ensure reproducibility, we release our code, preprocessed datasets, and model checkpoints with detailed documentation and scripts to replicate all experiments. PrivaMod is available at https://github.com/kvjunior/PrivaMod/blob/main/README.md .