Beyond the Rear Area: Integrating Supply Chain, Infrastructure, and Military Logistics Under Adversarial Threat
By LtCol M. Matthew Phelps, USMC
The commercial supply chain is now contested terrain. Logistics has always been contested at the tactical level — attacking convoys and targeting stockpiles are as old as war — but the reach of modern weapons extends well beyond what planners once called the rear area. Kinetic and non-kinetic means alike, from drone-delivered munitions to offensive cyber capabilities, can now touch the civilian systems that underpin military capability: raw material production, domestic manufacturing, and the transportation networks that move goods to the force. Yet there are no tools designed for this environment that are available to military planners.
Commercial supply chains employ supply chain risk management (SCRM) to prepare for stochastic fluctuations in supply and demand — port congestion, demand shocks, and pandemics. Critical infrastructure protection (CIP) planning safeguards physical networks against extreme weather and isolated attacks, optimizing for recovery and flow resumption. Military logistics planning protects the organic force — convoy escorts, facility hardening — and optimizes for resilience and survivability within the joint logistics enterprise. Each framework is appropriate to its own layer: the commercial supply chain, the physical infrastructure network, and the military logistics enterprise. None extends usefully into the others. They address different risk types and optimize for different variables.
The result is a structural blind spot.
When an adversary disrupts a rail node, port crane control system, or raw-material supplier, the cascading effects can ripple through commercial and civilian infrastructure into military sustainment — and remain invisible to existing planning tools.
Military logistics planning optimizes how to use available resources. It has no native mechanism for modeling how deliberate disruption of civilian systems degrades those resources before they ever reach the force.
Three Connected Failures
Three structural failures recur across the supply chain, network science, military logistics, and modeling literatures:
No multi-layer integration. Supply chain tools model flows of materials, cash, and information without mapping them to physical chokepoints. Network science rigorously models physical infrastructure — ports, rail, roads — but assesses node criticality through structural metrics such as betweenness centrality that are agnostic about operational significance. Military logistics doctrine addresses operational requirements without tracing disruptions backward through infrastructure into the industrial base. Most raw materials and industrial inputs for military goods originate in the commercial sector, yet the tools available for defense industrial base (DIB) planning stop at the boundary of the organic force.
No adversarial intent. Existing frameworks treat disruption as something that happens to a system — weather, demand shocks, random node failures. Where an attacker is considered at all, it is modeled as an abstract agent removing nodes by structural property. A rational adversary behaves differently: It observes dependencies and exploits them selectively, targeting the node whose removal produces the maximum operational effect at the very moment of greatest vulnerability — which may not be the most connected node. And the adversary adapts, observing friendly mitigation and retargeting.
No cross-layer cascade tracing. No existing framework traces how a disruption at a supplier or infrastructure node degrades tempo and reach at the operational layer. Even the most advanced contested-logistics methodologies — such as Powers and O’Flynn’s simulation-informed approximate dynamic programming (ADP) model, which generates near-optimal logistics decisions under port denial — treat upstream commercial and infrastructure conditions as fixed inputs rather than contested variables.
These failures are not independent deficiencies that can be fixed by bolting existing tools together. They are connected structurally. Because supply chain analysis does not map flows to physical nodes, it cannot identify the chokepoints an adversary would target. Because network science does not assess operational significance, it cannot tell planners which infrastructure failures matter. Because military doctrine does not trace disruptions backward through the industrial base, it cannot warn commanders before operational effects materialize. Closing one gap requires closing all three.
The Contested Sustainment Analysis Framework
The Contested Sustainment Analysis Framework (CSAF) is a requirements specification — not a built tool — defining the architecture a contested sustainment planning capability must satisfy.
The CSAF treats the entire network, from raw material to combat power, as a logistics battlespace that can be simulated and analyzed before committing real forces and resources.
The framework binds three layers — DIB supply chain, physical network infrastructure, and military operational logistics — by requiring each node to carry both a supply chain identity (what it produces or moves) and a physical identity (where it is and what threatens it), so that adversarial action on one layer propagates consequences through the others. A planning tool built on the CSAF must satisfy five core requirements.
Multi-layer integration. A disruption introduced at any layer must propagate through the interfaces to the others. A cyberattack on port crane control systems must register not only as an infrastructure disruption but as a reduction in supply chain throughput and as a constraint on operational reach and tempo.
Adversarial decision modeling. The adversary must be modeled as a goal-directed actor whose decisions respond to friendly adaptations — not as a stochastic disruption profile. Powers and O’Flynn demonstrate that Q-learning handles this adaptation loop without requiring prior knowledge of transition probabilities, a critical point because adversarial targeting decisions are not knowable in advance. The CSAF requires that this computational approach operates across all three layers simultaneously rather than within a single operational scenario.
Cascading effect propagation. Inter-layer propagation must be endogenous to the model’s architecture — built into its logic — rather than an external assumption the planner supplies by hand. A rail disruption must register simultaneously as a throughput constraint in the supply chain and a tempo constraint in the operational layer.
Operational output metrics. The tool must translate network performance into terms commanders use, such as tempo, reach, and available options. The relevant output is not “Port X operates at 70 percent capacity” but “the force can sustain combat operations at current intensity for N days before a critical munition shortfall forces a reduction in tempo.”
Adaptability and speed. Outputs must arrive fast enough to support decisions rather than follow them. In a peer competition where targeting decisions occur faster than planning cycles, computational architecture is a design requirement, not an implementation detail.
These requirements function as a system or not at all. Integration without cascade propagation produces a static map; cascade propagation without adversarial modeling produces stochastic rather than strategic analysis; adversarial modeling without operational metrics produces targeting insight commanders cannot act on.
Satisfying all five also produces a capability beyond their sum. Once nodes carry both identities and are linked across layers, the integrated map becomes an input for AI-assisted red-teaming — a systematic and adversarial framed traversal of the dependency graph, at a speed and scale no human planning team could match, searching not for the friendly force’s optimal logistics decision but for the adversary’s optimal disruption target.
A Cascade in Practice
To demonstrate both the framework’s analytical value and that red-teaming capability, I queried an AI language model with the CSAF’s requirements as the prompt and asked it to generate a scenario that existing frameworks would individually fail to identify. The following is a condensed narrative of the model’s output. It is entirely fictional.
A joint force conducting sustained combat operations depends on the continuous flow of a specific air defense interceptor — Munition X — to maintain coverage for forward operating bases. Munition X is produced by a single domestic manufacturer sourcing a specialty optical component from one of two qualified suppliers. Both suppliers rely on commercial rail to move components to the manufacturer, which ships completed rounds to an aerial port of embarkation for airlift to theater.
The adversary has mapped this dependency chain. Rather than striking hardened military targets, it launches a cyber intrusion against the industrial control systems of a major inland rail switching facility serving both suppliers — a facility that is not a military installation and sits outside force protection planning.
The cascade unfolds layer by layer. At the infrastructure layer, switching throughput drops roughly 60 percent for 18 to 22 days while operators restore manual protocols. One supplier, holding minimal inventory, suspends outbound shipments within 72 hours; the other reroutes to trucking, absorbing cost and a four-day transit delay. At the supply chain layer, the manufacturer — operating just-in-time, optimized for cost rather than surge — cannot sustain production; output of Munition X falls approximately 40 percent, and deliveries to the aerial port slip beginning on day 10. At the operational layer, theater stocks reach a critical threshold on day 28. The air component commander constrains air defense coverage to priority nodes, reducing ground maneuver options — and the adversary, anticipating this gap exactly, accelerates a planned offensive to exploit it.
From Framework to Prototype
Great power competition, lessons from Ukraine, and adversaries’ demonstrated willingness to target civilian infrastructure have effectively eliminated the rear area. The consequence is not merely tactical — it is analytical. Planners lack tools designed for this environment, and the tools they have are structurally incompatible with it.
The computational foundation for a CSAF-based tool exists: ADP methods have demonstrated rapid, near-optimal analysis at small-to-moderate scales without advance knowledge of adversarial behavior. The barriers are institutional and architectural — data integration across classification domains, alignment among the organizations that own each layer — and, therefore, are solvable with deliberate policy choices. The next logical step is a prototyping effort scoped to a single theater and a single class of adversarial action: narrow enough to be tractable, concrete enough to surface the data and authority gaps a general-purpose tool would require, and immediately valuable to the planners who use it.
The framework establishes what the tool must do. Building it is the work that remains. DTJ
About the Author
LtCol M. Matthew Phelps, USMC, is a Marine Corps logistics officer and a 2026 Distinguished Graduate of the Dwight D. Eisenhower School for National Security and Resource Strategy, National Defense University. He currently serves in the U.S. European Command J-7 Directorate.
This article is adapted from a research paper written at the Eisenhower School that was selected as the 2026 First Place Winner of the Lieutenant General Edward Honor NDTA Award. The views expressed are those of the author and do not reflect the official policy or position of the National Defense University, the Department of Defense, or the U.S. Government. To view the full paper in its entirety, go to https://www.ndtahq.com/wp-content/uploads/2026/08/LtCol-M.-Matthew-Phelps-Winner-4.pdf.
REFERENCES
1. Carol V. Evans, “Future Warfare: Weaponizing Critical Infrastructure,” Parameters 50, no. 2 (2020).
2. Matthew Powers and Brian O’Flynn, “Contested Logistics Simulation Output Analysis with Approximate Dynamic Programming,” Journal of Defense Analytics and Logistics 6, no. 2 (2022)
3. Oriol Artime et al., “Robustness and Resilience of Complex Networks,” Nature Reviews Physics 6, no. 2 (2024)
4. Réka Albert, Hawoong Jeong, and Albert-László Barabási, “Error and Attack Tolerance of Complex Networks,” Nature 406 (2000)
5. Joint Publication 4-0: Joint Logistics (2025)
AI disclosure: The author used Anthropic’s Claude to generate the fictional scenario described above, in response to a structured prompt containing the framework’s requirements, and to assist with editing. All analysis, interpretation, and argumentation are the author’s own.