SEI Proposes Five-Step Framework for 'Sufficient Realism' in Cyber Simulations

A digital twin raises an old design question

The post starts from a statement by the Army's chief of Network Command (NETCOM). In late August 2026, Maj. Gen. Jacqueline Denise McPhail said the Army needed a comprehensive simulation of its networks to take full advantage of artificial intelligence. According to Breaking Defense, she said such a simulation would create "an ultra-realistic training ground for network operators, cyber defenders, and AI algorithms on how the network works and how to keep it working, even when under attack."

Updyke says digital twins bring back a long-standing problem: how much of the operational environment actually needs to be reproduced. Every added system, behavior and dependency raises cost and complexity. A complete representation is expensive to build and hard to validate, and it is often unnecessary. Omitting the wrong element, however, can undermine an exercise even when everything else is modeled well.

The SEI has developed the GHOSTS framework since 2017. It began with simulated users operating real applications on hosts, producing the traffic, logs, files and browser history that defenders expect to see. It has since expanded into social and cognitive simulation, where actors communicate, form relationships, hold beliefs and influence one another. According to the SEI, GHOSTS models only the parts needed for a given experiment, exercise or mission rehearsal.

What 'sufficient realism' means

The SEI defines sufficient realism as "enough of the right factors in the environment, at the right fidelity, for the mission being rehearsed." It ties the idea to exercise validity. The first GHOSTS technical report, published in 2018, found that participants who rated an exercise as more realistic also rated the training as more valuable.

The concept builds on selective fidelity, the approach of the SIMNET Program. Instead of reproducing a tank in full physical detail, SIMNET started from the tasks soldiers had to perform. Designers could leave out a functional bilge-pump switch because it added nothing to platoon- or company-level tactical training. Seemingly minor cues such as seat rumble and visual movement stayed in, because crews used them to judge speed and terrain.

The post says cyber environments add a complication. Some elements must be realistic even if they are not part of the training objective, because they form the background against which mission-relevant activity is read. If all simulated users visit the same few sites, follow no deep links, produce almost no POST traffic and start work at the same instant, the background population becomes identifiable as synthetic. Once participants can tell simulation artifacts from ordinary activity, the exercise changes.

Taste as trained judgment

The post focuses on what it calls taste, a term for how experienced people make difficult choices when analysis does not give a single answer. Mission analysis can show which systems, behaviors, relationships and signals matter, and it can narrow the range of acceptable models. It cannot always say whether a behavior is believable enough, whether a discrepancy is consequential, or when more fidelity stops improving the rehearsal.

The SEI stresses that taste is not a substitute for mission analysis, measurement or validation. It applies only after the engineering problem has been narrowed. To answer the objection that a subjective judgment cannot be proven wrong, the post draws on Karl Popper's falsifiability as a standard of accountability. It also points to fields that rely on structured expert judgment, including RAND's Delphi method and the probabilistic risk assessments behind NUREG-1150, which covered accidents at five U.S. nuclear power plants.

For organizations planning cyber ranges or digital twins, the message is that the goal is not a perfect duplicate of reality. The question to ask is which parts of reality must hold up for the mission. The post is adapted from the SEI technical report A Decade of GHOSTS in the Machine.