- Mohammed Ahmed Abdullah Al-Anzi*
- Prince Sattam bin Abdulaziz University
- DOI: 10.5281/zenodo.18267556
Intrusion
Detection Systems (IDS) are critical for maintaining security in cloud
computing environments, where dynamic infrastructure and multi-tenancy present
unique challenges. This research implements and evaluates a machine
learning-based IDS specifically designed for Amazon Web Services (AWS) environments
using the CSE-CIC-IDS2018 dataset. Three machine learning algorithms—Isolation
Forest, One-Class Support Vector Machine (SVM), and Autoencoder neural
networks—were systematically compared based on standard performance metrics
including accuracy, precision, recall, F1-score, and Equal Error Rate (EER).
The Autoencoder model demonstrated superior performance with 96.8% accuracy and
3.3% EER, significantly outperforming traditional methods. Furthermore, we
propose a comprehensive AWS-native deployment architecture that integrates the
trained models with cloud services including Amazon SageMaker, Lambda,
CloudTrail, and Security Hub, creating a scalable, serverless IDS solution
capable of real-time threat detection and automated response. This study contributes
to the field of cloud security by providing both empirical validation of
machine learning approaches for anomaly detection and practical implementation
guidelines for AWS environments.

