Proactive AI-Based Cyber Deception for Cloud Threat Detection
Abstract
Sophisticated cyber adversaries increasingly target cloud computing infrastructures on Amazon Web Services (AWS). Traditional static security mechanisms often fail to counter dynamic attacks that exploit the elastic and distributed nature of cloud environments. This paper proposes an adaptive cyber deception framework for AWS augmented by artificial intelligence. The framework deploys realistic honeypots across EC2, S3, and Relational Database Service, supported by AWS Secrets Manager for secure credential handling, to emulate high-value assets. Real-time telemetry is analyzed using random cut forest and XGBoost models deployed through Amazon SageMaker to detect anomalous behavior. An adaptive orchestration layer using AWS Lambda dynamically modifies deception strategies, including elastic Internet Protocol (IP) rotation, S3 bait updates, and adaptive Relational Database Service schema enrichment. Experimental results show improved detection accuracy, reduced false positives, and effective adversarial engagement.