def get_payoff_mixed(model, attack_profile, defense_profile, attack_strategy,
defense_strategy):
"""
Function for computing the payoff of the defender given its mixed strategy and the mixed strategy of the attacker.
:param model: Model of the alert prioritization problem (i.e., Model object).
:param attack_profile: List of attack policies.ks given a model and a state.
:param defense_profile: List of defense policies.
:param attack_strategy: List of probablities of choosing policy from the attack profile
:param defense_strategy: List of probablities of choosing policy from the defense profile
:return: The expected discounted reward.
"""
total_discount_reward = 0
attack_policies = np.random.choice(attack_profile,
MAX_EPISODES,
p=attack_strategy)
defense_policies = np.random.choice(defense_profile,
MAX_EPISODES,
p=defense_strategy)
initial_state = Model.State(model)
for i in range(MAX_EPISODES):
state = initial_state
episode_reward = 0.0
defense_policy = defense_policies[i]
attack_policy = attack_policies[i]
for j in range(MAX_STEPS):
next_state = model.next_state('old', state, defense_policy,
attack_policy)
loss = next_state.U - state.U
state = next_state
step_reward = -1.0 * loss
episode_reward += GAMMA**j * step_reward
total_discount_reward += episode_reward
ave_discount_reward = total_discount_reward / MAX_EPISODES
return ave_discount_reward
if model.adv_budget == 3:
alpha = [1, 0, 1]
elif model.def_budget == 30:
if model.adv_budget == 2:
alpha = [0, 0, 1]
if model.adv_budget == 3:
alpha = [0, 0.33, 1]
return alpha
def test_attack_ids(model, state):
#alpha = [0, 1, 0.54, 1, 0, 0, 0]
alpha = [0, 0, 1, 1, 0, 0, 0.419]
return alpha
if __name__ == "__main__":
model = test_model_fraud(10, 2)
state = Model.State(model)
#print(test_defense_action(model, state))
#print(test_attack_action(model, state))
i = 0
while i < 10:
print('#############################')
print(i)
print('state:', state)
print('attacker:', test_attack_action(model, state))
state = model.next_state('old', state, test_defense_action,
test_attack_action)
i += 1
