Merge pull request #269 from stratosphereips/fix-reward-conf

Fix the names of the rewards in the configuration

Author: ondrej-lukas

AIDojoCoordinator/coordinator.py

@@ -206,7 +206,8 @@ async def _fetch_initialization_objects(self):
                         self.logger.error(f"Failed to fetch initialization objects. Status: {response.status}")
             except Exception as e:
                self.logger.error(f"Error fetching initialization objects: {e}")
-
+        # Temporary fix
+        self.task_config = ConfigParser(self._task_config_file)
     def _load_initialization_objects(self)->None:
         """
         Loads task configuration from a local file.
@@ -330,8 +331,8 @@ async def start_tasks(self):
         else:
             self._global_defender = None
         self._use_dynamic_ips = self.task_config.get_use_dynamic_addresses()
-        self._rewards = self.task_config.get_rewards(["step", "sucess", "fail"])
-        self.logger.debug(f"Rewards set to:{self._rewards}")
+        self._rewards = self.task_config.get_rewards(["step", "success", "fail"])
+        self.logger.info(f"Rewards set to:{self._rewards}")
 
         # start server for agent communication
         self._spawn_task(self.start_tcp_server)
@@ -587,7 +588,7 @@ async def _assign_rewards_episode_end(self):
                 for agent in attackers:
                     self.logger.debug(f"Processing reward for agent {agent}")
                     if self._agent_status[agent] is AgentStatus.Success:
-                        self._agent_rewards[agent] += self._rewards["sucess"]
+                        self._agent_rewards[agent] += self._rewards["success"]
                         successful_attack = True
                     else:
                         self._agent_rewards[agent] += self._rewards["fail"]
@@ -596,7 +597,7 @@ async def _assign_rewards_episode_end(self):
                 for agent in defenders:
                     self.logger.debug(f"Processing reward for agent {agent}")
                     if not successful_attack:
-                        self._agent_rewards[agent] += self._rewards["sucess"]
+                        self._agent_rewards[agent] += self._rewards["success"]
                         self._agent_status[agent] = AgentStatus.Success
                     else:
                         self._agent_rewards[agent] += self._rewards["fail"]

AIDojoCoordinator/netsecenv_conf.yaml

@@ -4,7 +4,7 @@ coordinator:
   # Define the conf for all the 'attacker' agents.
   agents:
     Attacker:
-      max_steps: 20
+      max_steps: 25
       goal:
         description: "Exfiltrate data to '213.47.23.195'"
         is_any_part_of_goal_random: True
@@ -103,9 +103,9 @@ env:
   use_firewall: True
   save_trajectories: False
   rewards:
-    win: 100
+    success: 100
     step: -1
-    loss: -10
+    fail: -10
   actions:
     scan_network:
       prob_success: 1.0