Machine Learning Security (MLSec) Attack Workbench For Red Teaming Of Autonomous Assets

Artificial Intelligence (AI) and Machine Learning (ML) assets are an integral part of securing the USAF and USSF Assets Freedom of Action in, to, and from the Space Domain. These assets are a key component of the Air Force Research Laboratory (AFRL) Space Vehicles Directorate Simulations and Analysis Branch (RVES) mission. RVES is responsible for Space Superiority Concepts to include Artificial Intelligence (AI) and Machine Learning (ML) modeling for space and multi-domain operations.  RVES investigates and develops solutions for trusted, high-performance systems that are necessary to ensure the survivability of key USSF and USAF systems. The RVES space domain modeling system is used to train operators and cyber defenders on attack, launch, and other critical aspects of the space domain battlespace. Adversaries like China and Russia are now using AI/ML to conduct cyber-attacks on AI/ML systems and assets that allow freedom of maneuver in the space domain. HiddenLayers MLSec Machine Learning Detection and Response (MLDR) and MLSec Attack Workbench will provide RVES with the research, development and execution plan necessary to identify and patch cyber vulnerabilities in their autonomous space assets and machine learning models. The HiddenLayer MLSec Attack Workbench, once adapted for USSF use, will allow AFRL/RVES to identify autonomous model vulnerabilities and demonstrate reasonable and feasible attacks sequences on currently identified and deployable models in a simulation realm to validate vulnerabilities prior to being placed on orbit, securing freedom of maneuver to, in, and from space. As we integrate machine learning and AI for decision-making, the HiddenLayer solutions will allow for adoption of trustworthy AI solutions for defense of the high frontier.