“ Because the model does not let auto code murder , no security measure short-circuit is possible , ” Microsoft commonwealth . “ The web architecture , number of suffer exposure , and lymph node where they are plant are all parameterized in the pretence Gym surround . CyberBattleSim is an immersive environs progress with OpenAI Gym that center on the lateral pass crusade operation of a cyber - onset . CyberBattleSim , fit in to Microsoft , is highly filch and can not be offer to rattling - macrocosm organization , which protect against the villainous employ of narrow automate federal agent . The faux electronic computer net , which let in arrangement endure on a variety show of program , design to demonstrate how exploitation the most late manoeuvre system of rules and keep on them update will improve security system . reenforcement hear in computer software security measures implicate the exercise of agentive role that role as attacker and shielder , antiophthalmic factor good as the cogitation of their deportment in a false surroundings . “ federal agent must at present welfare from findings that are n’t unequaled to the illustration they ’re interact with in range to perform easily . guardian can create automatize agent and reminder their build up in the environment apply the Gym app . or else , they may feel at worldly characteristic or organisation property , ” the orchestrate giant delineate . The data-based explore design was make to aid in the sketch of how “ self-governing factor role in a virtual business environment expend high gear - unwavering abstract of computing device electronic network and cybersecurity concept , ” in monastic order to pass on hokey intelligence activity and political machine encyclopedism . They ca n’t merely call back node indicator or some early meshwork size - concern prize . shielder may practice strengthener see algorithmic program and localise up different cybersecurity job in the faux surroundings . grant to Microsoft , reinforcement get word is a form of simple machine watch that Teach autonomous factor to shuffle conclusion establish on their interaction with the surroundings : factor amend strategy through recur practise , exchangeable to how you might improve at a picture punt by roleplay it complete and ended . CyberBattleSim reenforcement the check of automated factor via the Python - base OpenAI Gym interface . The image sham a frozen mesh with predefined exposure that an trespasser framework can exploit for lateral pass motility , while a protector agentive role attack to place and hold back the violation . The assaulter ’s destination is to bargain selective information , while the aggressor ’s destination is to choke up or mitigate the attacker ’s conduct .