Privacy-Aware Distributed Anomaly Detection Using Federated Analytics for Securing Multi-Tenant Cloud Environments

Abstract

The increasing adoption of multi-tenant cloud computing has introduced significant security challenges associated with data privacy, resource isolation, unauthorized access, and sophisticated cyber threats. Traditional centralized anomaly detection approaches require organizations to transfer sensitive operational data to centralized monitoring systems, potentially exposing confidential information and increasing privacy risks. This research proposes a privacy-aware distributed anomaly detection framework using federated analytics to strengthen security across multi-tenant cloud environments. The proposed framework enables multiple tenants to collaboratively train anomaly detection models without directly sharing their sensitive datasets. Federated learning techniques facilitate decentralized model training, while privacy-preserving mechanisms, including differential privacy, secure aggregation, and encrypted communication, protect tenant-specific information. The framework integrates distributed behavioral monitoring, machine learning-based anomaly identification, and collaborative threat intelligence to detect unusual access patterns, malicious activities, resource misuse, and potential security breaches. The research methodology considers heterogeneous tenant workloads, distributed data characteristics, and evolving cyberattack patterns to evaluate detection accuracy, privacy preservation, communication efficiency, and computational overhead. Performance evaluation is designed around established machine learning metrics, including precision, recall, F1-score, and detection latency. The proposed approach aims to provide scalable, intelligent, and privacy-conscious security monitoring while maintaining tenant autonomy and improving collaborative threat detection in dynamic cloud infrastructures.