Privacy-Aware Enterprise Cloud Intrusion Detection Through Generative AI and Federated Learning Architectures

Abstract

The increasing adoption of enterprise cloud computing has introduced significant cybersecurity challenges, particularly concerning sophisticated intrusion attempts, data privacy, and distributed threat detection. Traditional intrusion detection systems often depend on centralized data collection, which creates privacy risks, communication overhead, and limitations in identifying emerging cyber threats across heterogeneous cloud environments. This research proposes a privacy-aware enterprise cloud intrusion detection architecture integrating Generative Artificial Intelligence (GenAI) and Federated Learning (FL) to improve threat identification while maintaining data confidentiality. The proposed framework enables multiple enterprise cloud environments to collaboratively train intelligent intrusion detection models without exchanging sensitive raw network traffic or organizational information. Federated learning supports decentralized model development, while generative AI enhances threat analysis through synthetic security data generation, anomaly interpretation, and contextual threat classification. Privacy-enhancing mechanisms, including secure aggregation, differential privacy, and encrypted communication, are incorporated to strengthen information protection. The research methodology adopts a conceptual and experimental evaluation approach using representative network intrusion datasets and simulated distributed enterprise environments. Model performance is assessed through detection accuracy, precision, recall, F1-score, false-positive rate, communication overhead, and privacy considerations. The proposed architecture aims to establish a scalable, adaptive, and privacy-preserving cybersecurity framework capable of supporting intelligent intrusion detection across modern enterprise cloud infrastructures.