Loss Scenarios LS1—Timing issue: An adversary induces denial-of-service effects so data arrives too late, leading to a manipulated decision From a cybersecurity perspective, this loss scenario could be exploited through API delays, induced network latency, stale data, or weak time - stamp validation, and then tracked as a formal risk item for mitigation. Integrating MITRE ATLAS Integrating MITRE ATLAS into loss- based engineering provides a struc - tured way to connect AI-specific threat intelligence with system-level hazard and risk analysis. By incorpo - rating ATLAS tactics, techniques, and attack patterns into engineering work - flows, organizations can better iden - tify how AI-enabled systems may be manipulated or produce unsafe out - comes under adversarial conditions. Uncovering Hidden Risks A key outcome of the CRRM activity is its ability to identify actionable risks and emergent vulnerabilities that conventional testing methods often overlook, as illustrated by Loss Sce - nario 1. That timing-related loss sce - nario shows how interactions among data timing, system synchronization, and control logic can create vulner - abilities that may remain undetected in traditional validation processes. Conclusion As AI continues to reshape war- fighting acquisition and system de - sign, organizations need methods that address security, resilience, and mission assurance at the system
level. This need aligns directly with the department’s Acquisition Trans - formation Strategy, which calls on the enterprise to modernize systems engineering and to build and sustain weapon systems better, faster, and cheaper through advanced acquisi - tion and physics-based models. Extending the CRRM process into a digital, model-based environment operationalizes that intent. It embeds loss-based engineering into the same authoritative models used for acqui - sition decisions, shifting assurance “left” so that resilience is engineered- in rather than bolted-on after the fact. Continuing the CRRM process to software-intensive and AI-enabled systems within an MBSE environment offers a practical way to identify and mitigate complex risks earlier in the life cycle. The result is a more proac - tive, engineering-driven approach for uncovering hidden risks, improving system reliability, and strengthening
the security posture of next-genera - tion AI-enabled capabilities. Sources and Suggested Reading • MITRE. Adversarial Threat Landscape for Artificial-Intelligence Systems (ATLAS). • Clifford, M. SERC Systems and Cyber Resilience Modeling. Feb. 7, 2022. • Leveson, N. Engineering a Safer World: Systems Thinking Applied to Safety. MIT Press. • Young, W., & Leveson, N. STPA-Sec: Extending STPA for Security Analysis. • Google. Site Reliability Engineering: How Google Runs Production Systems. • Department of Defense. Acquisition Tr- ansformation Strategy . November 2025. JACOBSON is a professor of Cybersecu- rity and System Security Engineering at the Warfighting Acquisition University, where he leads the advancement of Secure Cyber Resilient Engineer and teaches hands-on Cyber Physical Training Ranges and MBSE Resiliency courses. He holds a doctorate in Information Assurance and has extensive experience across the DoW services in sys- tem security engineering, SCRM, cybersecu- rity, innovation, and software engineering. The author can be contacted through aaron.jacobson@waru.edu. The views expressed in this article are those of the author alone and not the Department of War. Reproduction or reposting of articles from Defense Acquisition magazine should credit the authors and the magazine.
To illustrate the methodology, consider a notional Automated AI Response Operating Node that performs alert triage, isolates affected systems, and recommends remediation actions. The use case diagram shows the system boundary, the external actors, and the primary operational functions.
40 DEFENSE ACQUISITION MAGAZINE | SEPTEMBER –OCTOBER 2026
Made with FlippingBook - Online Brochure Maker