Security Game

This environment is designed to testing any leader's (the protector) learning strategy in an unknown general-form security games, with unknown leader rewards and an unknown follower. At each round, the leader selects a certain number of targets to protect, followed by the follower’s response action j (column) determined by certain follower behavior mode. The leader/user then observes realized (possibly noisy) reward for action profile, which reflects if the follower has successfully launched an attack on an unprotected target.