A Robust Deep Learning Driven DDoS Defense Model for Cloud-Based Financial Infrastructures Using Variational Feature Encoding and Adaptive Attention Mechanisms

Abstract: Cloud-based network infrastructures have become major targets for Distributed Denial of Service (DDoS) attacks, specifically in the financial systems. Current cloud-native DDoS risk management services are effective against volumetric attacks but often fail to detect complex, zero-day, and application-layer threats. This research presents Secure Cloud-Fin (SCF), an intelligent and cloud-aware framework designed to detect and mitigate DDoS attacks in financial systems. The model combines Variational Autoencoders (VAE) for deep feature learning, for adaptive pattern recognition we used Attention-Enriched Transfer Learning (AETL), and for precise decision-making- XGBoost fusion. This new idea provides accurate and real-time protection against both volumetric and stealthy application-layer attacks. The system -Secure Cloud Fin-tested using benchmark datasets such as CIC-DDoS2019, CIC-DDoS2020, CRCDDoS2022, and UNSW-NB15. Secure Cloud-Fin can be deployed flexibly across edge, cloud, or hybrid environments, making it suitable for modern financial infrastructures. The results show that the proposed approach outperforms traditional models like CNN, LSTM, and Random Forest. Its adaptive design and high precision make it an effective solution for evolving DDoS threats. Overall, Secure Cloud-Fin ensures secure, scalable, and continuous financial operations in cloud-based ecosystems Keywords: DDoS Detection, Cloud Aware, Financial Systems, XGBoost, zero-day, application layer attacks