yawning_titan.agents.simple_blue.SimpleBlue#

class yawning_titan.agents.simple_blue.SimpleBlue(n_machines=None)[source]#

Bases: object

An interface capable of training a reinforcement learning agent in the specific environments.

Methods

do_blue_action

Perform the chosen action.

nothing

Do nothing - a noop in other reinforcement learning papers.

patch_machines

Patch a target machine and reduce its vulnerability score.

recover_machines

Recover a compromised machine and resets its state to the initial state during environment creation.

patch_machines(action, machine_states)[source]#

Patch a target machine and reduce its vulnerability score.

This action reduces a target machines vulerability score by 0.2 per use to a lower threshold of 0.2.

Parameters:
  • action – The chosen action

  • machine_states – The current state of the env

recover_machines(action, machine_states, initial_states)[source]#

Recover a compromised machine and resets its state to the initial state during environment creation.

Parameters:
  • action – The chosen action

  • machine_states – The current state of the env

  • initial_states – The first state of the environment when initialised/reset

nothing(action)[source]#

Do nothing - a noop in other reinforcement learning papers.

do_blue_action(action, machine_states, initial_states)[source]#

Perform the chosen action.

Parameters:
  • action – The chosen action to perform

  • machine_states – The state of the current env

  • initial_states – The state of the env initially