Corresponding Author: Prashant Soni
Web3 applications provide decentralized and user-centric digital services through blockchain networks, smart contracts, and wallet-based identities. However, their dynamic transaction behavior, pseudonymous identities, and continuously changing access conditions introduce significant security challenges. Conventional access-control mechanisms primarily rely on static policies and predefined rules, limiting their ability to respond to evolving security risks. This paper proposes a context-aware risk scoring and adaptive access-control model for Web3 applications that integrates transaction context, wallet behavior, device/session information, and identity/credential status. These contextual features are normalized and combined using weighted risk scoring to classify access requests into low-risk, medium-risk, and high-risk levels. The proposed framework applies differentiated security actions, including normal authorization, step-up verification, and access denial, while continuously reassessing contextual changes. A reproducible simulation-based evaluation is conducted using legitimate, suspicious, and malicious access requests and compared with a rule-based baseline. On the held-out test set, the proposed model achieves 99.12% accuracy, 97.48% precision, 96.67% recall, and 97.07% F1-score, while maintaining a false-positive rate of 0.44%. The model also achieves an average decision latency of 12.66 ms, compared with 8.22 ms for the rule-based approach. Ablation analysis further demonstrates the contribution of the individual contextual feature categories to security performance. The results indicate that context-aware risk scoring can improve malicious-access detection and enable more granular adaptive security decisions for Web3 applications.