AI-Driven Intrusion Detection Systems for Multi-Tenant Cloud Platforms

Abstract

The proliferation of multi-tenant cloud platforms—where resources are shared across diverse users—has introduced complex cybersecurity challenges, particularly in intrusion detection. Traditional systems often struggle with isolating tenant-specific behaviors and identifying emerging threats. In response, this paper proposes an AI-Driven Intrusion Detection System (AI-IDS) tailored for multi-tenant cloud environments. It integrates deep learning techniques—specifically Convolutional Neural Networks (CNNs)—with Transformer-based models and federated learning to enhance detection accuracy, privacy, and adaptability. The CNN component extracts spatial features from network data, while the Transformer module captures temporal attack patterns using attention mechanisms. Federated learning (FL) enables collaborative model training across tenant domains while retaining data privacy. Experimental evaluations on realistic, tenant-simulated traffic datasets demonstrate that the hybrid AI-IDS achieves superior intrusion accuracy—exceeding 94% overall—and reduces false positives compared to CNN-only or Transformer-only models. Federated learning further preserves privacy without significantly degrading performance. Moreover, the system demonstrates resilience against adversarial tactics and provides scalable deployment across tenant clusters. These results affirm that our AI-IDS provides a robust, efficient, and privacy-conscious solution for the evolving landscape of multi-tenant cloud security.