Data for the Paper "Attack-Aware and Efficient Virtual Machine Placement via Multi-Agent Reinforcement Learning"

This repository contains the anonymized data accompanying the paper "Attack-Aware and Efficient Virtual Machine Placement via Multi-Agent Reinforcement Learning".The repository is organized as follows: attacker/: Logs of the attacker agent trained under different threat models and reinforcement learning (RL) algorithms. defender/: Logs of various defender studies, including RL algorithm comparison, deployment, replica, scalability, sensitivity, threat models, and trade-off analyses. environment_database/: Database of cloud providers collected via Shodan, including service and operating system distributions, used to generate the scenarios. gae/: Graph Autoencoder logs and trained model. classifiers/: Classifiers integrated into the environment to approximate vulnerability outcomes and isolation levels. scenarios/: Pickle files representing the network scenarios used for training and testing the agents.