Adaptive and Secure ETL for Defense Data Systems Using Federated Reinforcement Learning
Abstract: In increasingly complex defense environments, ensuring secure, adaptive, and compliant data processing has become essential. This paper presents the Adaptive Secure ETL based on Federated Reinforcement Learning (AS-ETL-FRL), a novel framework designed to handle sensitive Department of Defense (DoD) data. Integrating federated deep autoencoders for anomaly detection with reinforcement learning agents for dynamic optimization, the system offers secure extraction, transformation, and loading (ETL) pipelines. Tested on the IBM Cloud Console Anomaly Detection Dataset, the model demonstrated 98.5% accuracy in anomaly detection, minimized latency, and maintained compliance with NIST, FISMA, and CMMC standards. By avoiding raw data sharing and enabling real-time decision-making, the AS-ETL-FRL architecture contributes significantly to privacy-preserving, intelligent data governance for mission-critical systems.
Keywords: Federated Learning, Reinforcement Learning, DoD Data Management, Anomaly Detection, Explainable AI
How to Cite?: Manohar Reddy Sokkula, "Adaptive and Secure ETL for Defense Data Systems Using Federated Reinforcement Learning ", Volume 14 Issue 10, October 2025, International Journal of Science and Research (IJSR), Pages: 879-888, https://www.ijsr.net/getabstract.php?paperid=SR251016172047, DOI: https://dx.doi.org/10.21275/SR251016172047