Siege proposes to develop the ADHoC (Adversarial Machine Learning for the Hardening of Cyber Defenses) software capability which combines a novel implementation of adversarial machine learning with a high-fidelity cyber simulation and evaluation environment to harden DML. The proposed solution provides a process for pairing cyber-attacks with AML algorithms in order to defeat DML systems. Through the process of learning how to defeat DML systems, ADHoC will identify the precise strategies, techniques, and data that led to a successful outcome. This knowledge can then be used to construct new training sets or improved ML models that can be used to harden the DML system against similar attacks. Over multiple iterations, both the AML attacker and DML defender will improve and produce more secure systems. Siege believes that the unique combination of novel AML implementations along with a high fidelity cyber simulation and evaluation environment is innovative and will significantly raise the bar for machine learning based intrusion detection and prevention leading to more secure systems for the U.S. Army.