Foundations of Autonomous Adaptive Cyber Systems
Abstract
Autonomy in physical and cyber systems promises to revolutionize military operations. The ability of autonomous systems to execute at scales, scopes, and tempos exceeding those of humans and human-controlled systems will introduce entirely new types of military strategies and tactics, especially in highly contested physical and cyber environments. The development and automation of cyber strategies that are responsive to autonomous adversaries pose basic new technical challenges for cyber-security. The ongoing Adaptive Cyber-Defense (ACD) ARO MURI project has developed a range of new ACD techniques and methodologies for reasoning in an adaptive environment. Building on that work, this proposed additional effort would address fundamental scientific problems essential for US domination in future autonomous cyber conflicts. To this end, this project will develop foundational concepts for Autonomous Adaptive Cyber Systems (A2CS). Specific areas that the team will explore include: ¥ Game and control theory-based MTDs and ACDs for fully autonomous cyber operations; ¥ The extent to which autonomous cyber systems can be designed and operated in a framework that is significantly different from the human-based systems we now operate; ¥ On-line learning algorithms, including deep recurrent networks, sketch-based algorithms and reinforcement learning, for the kinds of situation awareness and decisions that autonomous cyber systems will require; ¥ Human understanding and control of highly distributed autonomous cyber defenses; ¥ Quantitative performance metrics for the above so that autonomous cyber defensive agents can reason about the situation and appropriate responses as well as allowing humans to assess and improve the autonomous system.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Oct 16, 2018
- Source ID
- W911NF1710486
Entities
People
- Sushil Jajodia
Organizations
- Army Contracting Command
- George Mason University
- United States Army