FEDERATED AI OBSERVABILITY IN CROSS-JURISDICTIONAL CASINOS: A PRIVACY-PRESERVING FRAMEWORK FOR COMPLIANCE AND THREAT DETECTION

Abstract

As multi-property casinos expand across international jurisdictions, they face increasing pressure to maintain observability, enforce compliance, and share intelligence, without violating data residency laws or exposing sensitive logs. Traditional observability tools rely on centralized data aggregation, which is infeasible in privacy-constrained environments and introduces compliance risks.

This paper proposes a federated AI observability framework tailored for cross-jurisdictional casino networks. By leveraging edge AI agents, federated learning, and confidential computing, the architecture enables collaborative anomaly detection and compliance signal inference without centralized log sharing. The framework was simulated across three regional casino systems using Azure Confidential VMs and decentralized model aggregation.

Results demonstrate a 52% improvement in threat detection accuracy, zero raw log transfers, and full alignment with region-specific data governance mandates. This work represents a foundational step toward privacy-preserving observability in regulated, distributed enterprise systems and lays the groundwork for future extensions, such as real-time remediation workflows and Large Language Models (LLM)-driven context enrichment for intelligent monitoring.