Adaptive trust evaluation model based on container security analysis in the Docker platform

Adaptive trust evaluation model based on container security analysis in the Docker platform

연구 분야: Software Development

학회: International Journal of Information Security

The growing adoption of containers, driven by their lightweight and efficient characteristics, has underscored the critical importance of addressing security challenges. However, existing trust evaluation models often exhibit limitations such as low accuracy, excessive computational overhead, and slow detection speeds, limiting their practical deployment in container security analysis. To overcome these shortcomings, this paper introduces an Adaptive Trust Evaluation Model for Docker Container Security (CS-ATEM). The proposed model integrates random forest classification with k-means clustering to dynamically assess the trustworthiness of containers. To facilitate real-time and precise threat detection, the model decomposes trust evaluation into two components: direct trust value and interactive trust value. An adaptive weighting mechanism based on entropy weight method is devised to evaluate container parameters, ensuring that the assessment adapts to the evolving behavior of containers. Subsequently, a comprehensive trust value is derived through weighted aggregation of these two metrics, offering a holistic representation of a container’s trust state. The results of the experiment show that our model achieves a 10% higher accuracy compared to BLTM and a 20% improvement over SWM in detecting malicious containers, while simultaneously reducing response times. These results highlight the effectiveness of CS-ATEM in providing robust and dynamic trust evaluation for container security in highly adaptable environments.