An Apache-Centric Explainable AI Framework for Real-Time Cloud Cybersecurity: Multimodal Threat Intelligence and Integrated Credit–Fraud Risk Modeling Using Multivariate Classification

Abstract

The increasing complexity of cloud ecosystems has amplified the demand for intelligent, transparent, and scalable security architectures capable of delivering real-time threat detection and financial risk mitigation. This paper proposes an Apache-centric Explainable AI (XAI) framework that integrates multimodal threat intelligence, multivariate classification, and credit–fraud risk modeling within a unified cloud-native environment. Leveraging Apache Kafka for high-throughput data streaming, Apache Spark for distributed analytics, and Apache Flink for low-latency event processing, the framework fuses heterogeneous data—network telemetry, user behavior logs, text-based indicators, transaction patterns, and financial risk signals—to construct a comprehensive threat and fraud intelligence pipeline. The core analytical layer employs multivariate classification models, including SHAP-enabled deep neural networks, interpretable ensemble learners, and hybrid multimodal classifiers that capture correlations across numerical, categorical, temporal, and text-based features. Explainability mechanisms provide transparent justifications for alerts, enabling analysts and auditors to understand causal factors contributing to cybersecurity intrusions, credit anomalies, and fraudulent activities.