Secure AI Processing Platform for Expendable Systems

When quickly deployed into the hands of the warfighter, smart, low-cost expendable platforms are becoming essential to military operations by providing an asymmetric effect on the battlefield. This asymmetric capability has been recently demonstrated in Ukraine by the success of loitering munitions like Switchblades and Phoenix Ghosts. Rapidly evolving technologies in these small, expendable systems pose challenges to developers as well as users including development time, cybersecurity threats, and the Size, Weight, Power, and Cost (SWaP-C) optimization. To reduce development time, Digital Engineering (DE) techniques are necessary to rapidly explore and refine possible solutions. Technologies developed using DE must be translated from simulation to reality on a rapid scale to advance the state of the art in autonomous systems and provide real-world validation. To produce the numbers needed, the SWaP-C of these systems must be minimized by leveraging low-cost commercial processing hardware. This optimization becomes more challenging as increased computational requirements of these systems are needed to host advanced Artificial Intelligence and Machine Learning (AI/ML) technologies. While open systems standards such as the Weapon Open Systems Architecture (WOSA) enable mixing “best of breed” software from different sources, AI/ML algorithms come from many sources with varying development pedigrees. As a result, this high value AI/ML software must be protected along with other Critical Program Information (CPI) and execute in a secure environment to isolate complex software components that may contain malicious code. These small, expendable systems need to react quickly to changing operational conditions by enabling rapid technology insertion without compromising cybersecurity. In the proposed program, the Secure AI Processing Platform for Expendable Systems (SAPPES) solution will be developed. The solution will enhance AFRL’s AFSIM with processor and AI hardware digital twin capabilities to enable direct transition of the code developed in simulation secure, low SWaP-C, embedded computing hardware. The SAPPES embedded software architecture will be based on Secmation’s Anneal architecture, in development for the Missile Defense Agency (MDA), integrated with Secmation’s SoCrypt, in development for AFRL RV, to provide a Secret and Below (SAB) encryption capability. The hardware will be based on the Xilinx Versal AI Edge System on Chip (SoC). This recently released COTS device provides advanced AI acceleration and comes from a processor family used in secure DoD applications. The Versal AI Edge hardware enhanced by the Secmation’s advanced, purpose-built DoD security architecture capable of supporting classified information will result in a Secure AI Processing Platform. The program will conclude with integration and demonstration of SAPPES on a seeker processor in the AFRL Munitions Open Architecture Test and Evaluation Lab (MOATEL).