Deep Reinforcement Learning for Accurate Anomaly Detection in Multi-Cloud Environment

Intrusion Detection System (IDS) is used for identifying complicated Cloud assaults. A cooperative IDS may use an aggregation technique to decide on suspicious intrusions. Aggregation methods and IDS feedback slow the process. These restrictions make real-time cooperative IDS choices useless and unsustainable. We suggest a machine learning-based collaborative IDS that uses past feedback data to make proactive decisions to address these issues. DRL-based multi-cloud IDS was suggested. The suggested approach uses a DE noising Auto encoder (DA) to create a deep neural network. DA can rebuild IDS feedback from incomplete feedback. The DA network uses Q-learning to continuously learn features. This lets us respond on suspected incursions without IDS input. The GPU-enabled Tensor Flow model was tested on a real-world dataset. Our model achieves 95% detection accuracy in experiments.