A.S. Sebyakin 1

1 Financial University under Russian Government (Moscow, Russia)
1 249702@edu.fa.ru

This article investigates the trusted use of generative video in digital twins of situation centers, where synthetic visualizations can support training, briefing, post incident analysis, and what if scenario exploration, yet may also introduce serious decision risks. The paper focuses on a core challenge: realistic video can mislead operators when it contains semantic hallucinations, spatial inconsistencies, implausible motion, or incorrect event order and duration. To address this problem, the study proposes a formal trust protocol built as a set of validation gates that assess extracted features, such as object trajectories and event timestamps, rather than raw pixels.

The generative video module is modeled as a stochastic mapping from structured scenario packages to possible video realizations, allowing multiple outputs for the same input conditions. On this basis, the article defines a taxonomy of failure modes relevant to decision grade synthetic video. These include geometric violations, kinematic anomalies, event temporal errors, and unsupported or missing objects and events. Validation gates check spatial admissibility, kinematic consistency, event logic, and expected cardinality. The paper also introduces gate coverage as the probability that at least one gate detects a given failure mode, and estimates this quantity through Monte Carlo simulation with confidence intervals.

To demonstrate the approach, the author develops an imitation benchmark operating on extracted features instead of full video frames. The benchmark enables programmable fault injection, including trajectory teleportation, drift beyond permitted zones, hallucinated objects, inversion of event order, and omission of critical events. Numerical experiments compare baseline and extended gate configurations and reveal a clear robustness sensitivity trade off. Quantile based statistical checks reduce false rejections under noisy conditions, while additional outlier gates substantially improve detection of rare but operationally critical jumps and abnormal accelerations, though they also increase the false positive rate.

The proposed protocol offers practical value for designing and governing digital twin software in situation centers. It supports threshold calibration, red team testing of synthetic video modules, regulated release of generated materials for training and analysis, and creation of auditable evidence bundles for trustworthy operational use. Overall, the study provides a reproducible framework for measuring validation performance and balancing sensitivity against robustness in high stakes synthetic visualization workflows safely.

Sebyakin A.S. A mathematical model and numerical evaluation of the trust protocol for generative video in digital twins of situation centers. Nonlinear World. 2026. V. 24. № 2. P. 42–49. DOI: https:// doi.org/10.18127/ j20700970-202602-05 (In Russian)

1. Sebyakin A.S. Obzor metodov iskusstvennogo intellekta dlya generacii videokontenta i ih primenenie v postroenii cifrovyh dvojnikov situacionnyh centrov.M.: Finansovyj universitet pri Pravitel'stve Rossijskoj Federacii. 2025 (In Russian).
2. Grieves M.,Vickers J. Digital twin: Mitigating unpredictable, undesirable emergent behavior in complex systems. In: Transdisciplinary perspectives on complex systems. Springer. 2017.
3. Maler O., Nickovic D. Monitoring temporal properties of continuous signals. In: Formal modeling and analysis of timed systems (FORMATS). 2004.
4. Donzé A., Maler O. Robust satisfaction of temporal logic over real-valued signals. In: Formal modeling and analysis of timed systems (FORMATS). 2010.
5. Breck E., Cai S., Nielsen E., Salib M., Sculley D. The ML test score: A rubric for ML production readiness and technical debt reduction. arXiv preprint arXiv:1711.07973, 2017.
6. Donzé A. Breach, a toolbox for verification and parameter synthesis of hybrid systems. In: Computer aided verification (CAV). 2010.
7. Mitchell M. et al. Model cards for model reporting. In: Proceedings of the conference on fairness, accountability, and transparency (FAT*). 2019.
8. Gebru T. et al. Datasheets for datasets. In: Communications of the ACM. 2018.
9. NIST. Artificial intelligence risk management framework (AI RMF 1.0). National Institute of Standards and Technology. 2023.
10. Wolf K. et al., Towards a digital twin for supporting multi-agency incident management in a smart city. Scientific Reports. 2022. V. 12. Art. 16221.
11. Bartocci E., Deshmukh J.V., Donzé A., Fainekos G.E., Maler O., Ničković D., Sankaranarayanan S. Specification-Based Monitoring of Cyber-Physical Systems: A Survey on Theory, Tools and Applications. In: Lectures on Runtime Verification, Springer, 2018, pp. 135–175.
12. Coalition for Content Provenance and Authenticity (C2PA). C2PA Technical Specification. Version 2.2. 2025.
13. Unterthiner T. et al. Towards Accurate Generative Models of Video: A New Metric & Challenges. arXiv:1812.01717. 2018.