Gated Recurrent Unit-Based Detection Framework for Digital Wallet Attacks in Serverless Computing Environments

Digital Wallet Attacks (DWAs) are low-frequency but high-impact cyber threats that exploit vulnerabilities in cloud-based and serverless infrastructures. Due to characteristics such as auto-scaling, pay-per-use billing, and limited control over resource allocation, Function-as-a-Service (FaaS) architectures are particularly vulnerable, making timely detection crucial to prevent financial loss and maintain system integrity. This paper introduces a Digital Wallet Attack Detection Framework, termed GRU-DWAD, developed as part of a broader Cloud Security Assurance Framework (CSAF). The proposed framework operates in three stages: data pre-processing using Z-score normalization to ensure consistency, feature selection through Recursive Feature Elimination (RFE) to retain relevant attributes, and prediction using a fine-tuned Stacked Long Short-Term Memory (SLSTM) model for accurate detection. Extensive experiments on large-scale benchmark datasets demonstrate that GRU-DWAD achieves a detection accuracy of 95.30%, outperforming existing state-of-the-art techniques. The findings highlight the framework’s robustness, scalability, and explainability in combating price-based cyberattacks in serverless financial ecosystems, offering a reliable AI-driven solution for enhancing the security of digital wallets in modern cloud environments.