An open-source cyber range for training and evaluating reinforcement learning agents on network security problems.
Emulate large-scale IT networks and execute Markov games using OpenAI Gym-compatible Python environments.
CSLE is a research platform designed for running reinforcement learning experiments on cybersecurity control problems. It combines Linux container emulation with Markov decision processes and Markov games to test automated security policies. You should use it if you design, simulate, or benchmark autonomous security agents and AI defenders.
CSLE provides a multi-system architecture featuring Linux container network emulation, Gym-compatible simulation environments, and a management stack with Web UI, CLI, and REST APIs. It lets you model cyber attacks, client populations, and defensive actions using computational game theory and dynamic programming. You can monitor simulation traces in real time or hook up custom Python reinforcement learning libraries.
This tool suits cybersecurity researchers, AI developers, and network engineers building autonomous defense models or Markov game simulations. It is not meant for general game developers or teams seeking out-of-the-box, ready-made commercial security scanners. If you lack infrastructure to run Linux containers or compute clusters, this tool will be difficult to operate.
Simulate large-scale networks, cyber attacks, and client populations using distributed Linux containers.
Run Markov decision processes and Markov games directly using standard OpenAI Gym and Gymnasium interfaces.
Control and monitor emulation and simulation runs via a Web UI, CLI, REST API, or Python SDK.
Combine reinforcement learning with computational game theory, causal inference, dynamic programming, and evolutionary algorithms.
CSLE is free to access on GitHub, though no explicit license file is included in the main repository metadata. The primary financial cost comes from running and maintaining the server hardware or distributed compute cluster required for full-scale network emulations.
896 open remote AI-training roles, updated November 16, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Yes, but local laptop setups are restricted to small emulation topologies due to container resource limits.
Yes, its simulation system adheres to OpenAI Gym and Gymnasium standards so you can plug in custom RL algorithms.
You can manage and monitor simulation or emulation runs through the REST API, Web UI, CLI, or Python client libraries.
See Zekai first in Google