Optimized Isolation Forest based Docker Container Anomaly Monitoring System

This paper presents an enhanced anomaly detection system for Docker containers, leveraging an improved version of the Isolation Forest algorithm. Due to the highly dynamic and ephemeral nature of containerized environments, traditional monitoring approaches often fall short in identifying issues such as performance degradation, security breaches, and inefficient resource usage. Docker containers are widely adopted for deploying scalable applications in isolated environments, but their complexity makes consistent and accurate monitoring a challenge. To address this, our approach integrates hyperparameter tuning and optimized feature selection into the Isolation Forest model, significantly improving its ability to detect abnormal behavior. We evaluate the system using real-world Docker container data, and the results demonstrate superior performance compared to conventional methods. The proposed solution effectively identifies both known and previously unseen anomalies in realtime, without relying on labeled datasets.