YAWNING-TITAN v2.0.1 releases: abstract, graph based cyber-security simulation environment

YAWNING-TITAN

YAWNING-TITAN (YT) is an abstract, graph based cyber-security simulation environment that supports the training of intelligent agents for autonomous cyber operations. YAWNING-TITAN currently only supports defensive autonomous agents who face off against probabilistic red agents.

YT contains a small number of specific, self-contained OpenAI Gym environments for autonomous cyber defense research, which are great for learning and debugging, as well as a flexible, highly configurable generic environment which can be used to represent a range of scenarios of increasing complexity and scale. The generic environment only needs a network topology and a settings file in order to create an OpenAI Gym compliant environment which also enables open research and enhanced reproducibility.

YT has been designed with the following things in mind:

• Simplicity over complexity
• Minimal Hardware Requirements
• Operating System agnostic
• Support for a wide range of algorithms
• Enhanced agent/policy evaluation support
• Flexible environment and game rule configuration
• Generation of evaluation episode visualisations (gifs)

What is YAWNING-TITAN built with

YT is built on the shoulders of giants and heavily relies on the following libraries:

• OpenAI’s Gym is used as the basis for all of the environments
• Networkx is used as the underlying data structure used for all environments
• Stable Baselines 3 is used as a source of RL algorithms
• Rllib (part of Ray) is used as another source of RL algorithms

Changelog v2.0.1

What’s New

GUI

• Added new options in the Run page for separate generation of Animated GIFs and WEBM videos
  • This can be found under the Evaluation tab
    
  • If both Animated GIF and WEBM videos are to be generated, the video will be shown as the preview

Python API

• Separated saving of Animated GIFs and WEBM videos, so that they can be generated independently of each other
  

🐛 Bug Fixes

• Fixed an issue where an error occurs if no webm_output_directory is provided when the ActionLoop.gif_action_loop() is called
• WEBM files are now generated in the USER_DIR/yawning_titan/videos by default, but can be overridden by providing a path to webm_output_directory

Install & Use

Crown Copyright 2022 Defence Science and Technology Laboratory UK