901 - merged with dev

src/primaite/common/enums.py

@@ -1,7 +1,7 @@
 # Crown Copyright (C) Dstl 2022. DEFCON 703. Shared in confidence.
 """Enumerations for APE."""
 
-from enum import Enum
+from enum import Enum, IntEnum
 
 
 class NodeType(Enum):
@@ -32,6 +32,7 @@ class Priority(Enum):
 class HardwareState(Enum):
     """Node hardware state enumeration."""
 
+    NONE = 0
     ON = 1
     OFF = 2
     RESETTING = 3
@@ -42,6 +43,7 @@ class HardwareState(Enum):
 class SoftwareState(Enum):
     """Software or Service state enumeration."""
 
+    NONE = 0
     GOOD = 1
     PATCHING = 2
     COMPROMISED = 3
@@ -79,6 +81,65 @@ class Protocol(Enum):
     NONE = 7
 
 
+class SessionType(Enum):
+    """The type of PrimAITE Session to be run."""
+
+    TRAIN = 1
+    "Train an agent"
+    EVAL = 2
+    "Evaluate an agent"
+    TRAIN_EVAL = 3
+    "Train then evaluate an agent"
+
+
+class AgentFramework(Enum):
+    """The agent algorithm framework/package."""
+
+    CUSTOM = 0
+    "Custom Agent"
+    SB3 = 1
+    "Stable Baselines3"
+    RLLIB = 2
+    "Ray RLlib"
+
+
+class DeepLearningFramework(Enum):
+    """The deep learning framework."""
+
+    TF = "tf"
+    "Tensorflow"
+    TF2 = "tf2"
+    "Tensorflow 2.x"
+    TORCH = "torch"
+    "PyTorch"
+
+
+class AgentIdentifier(Enum):
+    """The Red Agent algo/class."""
+
+    A2C = 1
+    "Advantage Actor Critic"
+    PPO = 2
+    "Proximal Policy Optimization"
+    HARDCODED = 3
+    "The Hardcoded agents"
+    DO_NOTHING = 4
+    "The DoNothing agents"
+    RANDOM = 5
+    "The RandomAgent"
+    DUMMY = 6
+    "The DummyAgent"
+
+
+class HardCodedAgentView(Enum):
+    """The view the deterministic hard-coded agent has of the environment."""
+
+    BASIC = 1
+    "The current observation space only"
+    FULL = 2
+    "Full environment view with actions taken and reward feedback"
+
+
 class ActionType(Enum):
     """Action type enumeration."""
 
@@ -130,6 +191,14 @@ class LinkStatus(Enum):
     OVERLOAD = 4
 
 
+class SB3OutputVerboseLevel(IntEnum):
+    """The Stable Baselines3 learn/eval output verbosity level."""
+
+    NONE = 0
+    INFO = 1
+    DEBUG = 2
+
+
 class RulePermissionType(Enum):
     """Any firewall rule type."""
 

src/primaite/common/training_config.py

@@ -1,95 +0,0 @@
-# # Crown Copyright (C) Dstl 2022. DEFCON 703. Shared in confidence.
-# """The config class."""
-# from dataclasses import dataclass
-#
-# from primaite.common.enums import ActionType
-#
-#
-# @dataclass()
-# class TrainingConfig:
-#     """Class to hold main config values."""
-#
-#     # Generic
-#     agent_identifier: str  # The Red Agent algo/class to be used
-#     action_type: ActionType  # type of action to use (NODE/ACL/ANY)
-#     num_episodes: int  # number of episodes to train over
-#     num_steps: int  # number of steps in an episode
-#     time_delay: int  # delay between steps (ms) - applies to generic agents only
-#     # file
-#     session_type: str  # the session type to run (TRAINING or EVALUATION)
-#     load_agent: str  # Determine whether to load an agent from file
-#     agent_load_file: str  # File path and file name of agent if you're loading one in
-#
-#     # Environment
-#     observation_space_high_value: int  # The high value for the observation space
-#
-#     # Reward values
-#     # Generic
-#     all_ok: int
-#     # Node Hardware State
-#     off_should_be_on: int
-#     off_should_be_resetting: int
-#     on_should_be_off: int
-#     on_should_be_resetting: int
-#     resetting_should_be_on: int
-#     resetting_should_be_off: int
-#     resetting: int
-#     # Node Software or Service State
-#     good_should_be_patching: int
-#     good_should_be_compromised: int
-#     good_should_be_overwhelmed: int
-#     patching_should_be_good: int
-#     patching_should_be_compromised: int
-#     patching_should_be_overwhelmed: int
-#     patching: int
-#     compromised_should_be_good: int
-#     compromised_should_be_patching: int
-#     compromised_should_be_overwhelmed: int
-#     compromised: int
-#     overwhelmed_should_be_good: int
-#     overwhelmed_should_be_patching: int
-#     overwhelmed_should_be_compromised: int
-#     overwhelmed: int
-#     # Node File System State
-#     good_should_be_repairing: int
-#     good_should_be_restoring: int
-#     good_should_be_corrupt: int
-#     good_should_be_destroyed: int
-#     repairing_should_be_good: int
-#     repairing_should_be_restoring: int
-#     repairing_should_be_corrupt: int
-#     repairing_should_be_destroyed: int  # Repairing does not fix destroyed state - you need to restore
-#
-#     repairing: int
-#     restoring_should_be_good: int
-#     restoring_should_be_repairing: int
-#     restoring_should_be_corrupt: int  # Not the optimal method (as repair will fix corruption)
-#
-#     restoring_should_be_destroyed: int
-#     restoring: int
-#     corrupt_should_be_good: int
-#     corrupt_should_be_repairing: int
-#     corrupt_should_be_restoring: int
-#     corrupt_should_be_destroyed: int
-#     corrupt: int
-#     destroyed_should_be_good: int
-#     destroyed_should_be_repairing: int
-#     destroyed_should_be_restoring: int
-#     destroyed_should_be_corrupt: int
-#     destroyed: int
-#     scanning: int
-#     # IER status
-#     red_ier_running: int
-#     green_ier_blocked: int
-#
-#     # Patching / Reset
-#     os_patching_duration: int  # The time taken to patch the OS
-#     node_reset_duration: int  # The time taken to reset a node (hardware)
-#     node_booting_duration = 0  # The Time taken to turn on the node
-#     node_shutdown_duration = 0  # The time taken to turn off the node
-#     service_patching_duration: int  # The time taken to patch a service
-#     file_system_repairing_limit: int  # The time take to repair a file
-#     file_system_restoring_limit: int  # The time take to restore a file
-#     file_system_scanning_limit: int  # The time taken to scan the file system
-#     # Patching / Reset
-#