Adversarial Data Modeling in Epidemiology

Yiqi Su; Christo Kurisummoottil Thomas; Walid Saad; Sanmay Das; Bud Mishra; Naren Ramakrishnan

cs.GTFeb 23, 2026

Adversarial Data Modeling in Epidemiology

Authors: Yiqi Su, Christo Kurisummoottil Thomas, Walid Saad, Sanmay Das, Bud Mishra, Naren Ramakrishnan

Organizations: Virginia Tech Alexandria, VA, USA · Worcester Polytechnic Institute Worcester, MA, USA · New York University New York, NY, USA

Abstract

Epidemiological models increasingly rely on crowdsourced, self-reported behavioral data such as vaccination status, mask usage, and social distancing adherence. This data, however, is not passively sampled but instead strategically reported, making it a canonical case of adversarial input to a data mining pipeline. Individuals misreport for various reasons, e.g., to avoid penalties, to access benefits, or to express distrust in public health authorities. We introduce a data-modeling framework that casts the interaction between the population and a public health authority as a signaling game. This approach provides both a generative model of strategically-corrupted behavioral data and a mechanism for the receiver to recover reliable signal from it. Individuals (senders) choose how to report their behaviors, while the public health authority (receiver) updates their epidemiological model(s) based on potentially distorted signals, and modifies its trust in incoming reports accordingly. Focusing on deception around masking and vaccination, we characterize analytically game equilibrium outcomes as distinct regimes of data corruption, and evaluate the degree to which deception can be tolerated while maintaining epidemic control through policy interventions. In large scale simulations, our results show that even under pervasive dishonesty in pooling equilibria, well-designed sender and receiver strategies can still maintain effective epidemic control. Real-world validation further shows that behavioral distortions often exhibit structured patterns rather than arbitrary noise. This work advances the understanding of adversarial data in epidemiology and offers tools for designing more robust public health models in the presence of strategic user behavior.

Figures & tables

Figure 1. Illustration of the signaling game framework.

Figure 2. Compartmental SVEAIR epidemiological model with vaccination and masking. The ‘H’ is not a separate compartment but denotes (observed) hospitalization numbers, typically a fraction of the true numbers from the infected compartment.

Figure 3. Controlled reproduction number ( $R_{c}$ ) across equilibria vs no-interaction baseline with low (a) and high (b) initial behavioral rates.

Figure 4. Sender strategies $g(m\mid\theta)$ and receiver beliefs $q(\theta\mid m)$ across equilibria.

Figure 5. The deception rates across equilibria under low initial behavioral rates.

Figure 6. Comparison of adaptive and random policies under partial pooling with low initial behavioral rates, across epidemic dynamics, policy actions, signaling behavior, and utilities.

Figure 7. Maximum tolerable active deception under partial pooling across epidemic stages.

	Disease Control Score ( $\Delta$ )			Peak Hospitalization (ratio)
Factor	Sep.	Partial	Pooling	Sep.	Partial	Pooling
IHR	0.065	0.046	0.028	1.993	1.985	1.975
Incentives	-0.085	-0.065	-0.042	1.021	1.029	1.035
NRS	-0.263	-0.188	-0.117	1.120	1.144	1.163
Vaccine efficacy	0.325	0.320	0.243	0.989	0.981	0.972

Table 1. Stress-test outcomes across equilibria. Disease control score shows changes relative to baseline, while peak hospitalization reports ratios relative to baseline.

Appendix figures & tables8 assets

Supplementary material from the paper’s appendix.

Appendix

Type	True status	Vaccine report	Mask report
$T_{1}$	00	0 (Honest)	0 (Honest)
$T_{2}$	01	0 (Honest)	0 (Deceptive)
$T_{3}$	01	0 (Honest)	1 (Honest)
$T_{4}$	10	0 (Deceptive)	0 (Honest)
$T_{5}$	10	1 (Honest)	0 (Honest)
$T_{6}$	11	0 (Deceptive)	0 (Deceptive)

Appendix

Table A1 . Nine behavioral types defined in terms of how people with different compliance combinations report truthfully or deceive about their vaccination and/or masking status. (See A.2 for explanation.)

Parameter	Description	Value	References
Epidemiological model parameters
$\mu$	Natural death rate	$1/(75\times 365)$	( Longini et al., 2004 )
$\beta_{0}$	Baseline transmission rate	$0.35$	( Jing et al., 2021 )
$\delta$	Vaccine efficacy	$0.45$	( Dawood et al., 2020 )
$1/\gamma$	Infectious period (days)	$10$	( Walsh et al., 2020 )
$b$	Infectiousness of asymptomatic	$0.5$	( Longini et al., 2004 )

Appendix

Table A2 . Summary of model parameters and baseline values used in our experiments.

Figure A1 . The deception rates across equilibria under high initial behavioral rates.

Figure A2 . Comparison of adaptive and random policies under partial pooling with high initial behavioral rates, across epidemic dynamics, policy actions, signaling behavior, and utilities.

Figure A3 . Comparison of adaptive and random policies under separating with low initial behavioral rates, across epidemic dynamics, policy actions, signaling behavior, and utilities.

Figure A4 . Comparison of adaptive and random policies under separating with high initial behavioral rates, across epidemic dynamics, policy actions, signaling behavior, and utilities.

Figure A5 . Comparison of adaptive and random policies under pooling with low initial behavioral rates, across epidemic dynamics, policy actions, signaling behavior, and utilities.

Figure A6 . Comparison of adaptive and random policies under pooling with high initial behavioral rates, across epidemic dynamics, policy actions, signaling behavior, and utilities.

Corruption Robustness

Corruption robustness studies how to keep learning systems reliable when data, inputs, labels, or feedback are noisy, missing, or adversarially altered.

See all 50 papers

Toward World Models for Epidemiology
Apr 10, 2026Zeeshan Memon, Yiqi Su, Christo Kurisummoottil Thomas +3World Models Causal World Models
Department of Computer Science, Emory University, Atlanta, GA, USA · Department of Computer Science, Virginia Tech, Alexandria, VA, USA · Department of Electrical and Computer Engineering, Worcester Polytechnic Institute, Worcester, MA, USA +1
Integrating Persuasion Theory into the Epidemiological Modelling of Health Misinformation Spread on Social Media
Aug 16, 2026Mkululi Sikosana, Sean Maudsley-Barton, Oluwaseun AjaoDynamical Systems Social Media Analysis
Department of Computing and Mathematics, Manchester Metropolitan University, Manchester M15 6BH, United Kingdom
Generating Public Health Responses using Survey-Augmented Large Language Models
Jun 20, 2026Leonardo Marciaga, Thuyen Pham, Julia Rezvani +4Synthetic Data Augmentation Synthetic Data Generation
Illinois Institute of Technology · University of Massachusetts Amherst · Portland State University +4

Apr 10, 2026J/K move · Enter open · S save

Toward World Models for Epidemiology

Zeeshan Memon, Yiqi Su, Christo Kurisummoottil Thomas, Walid Saad+2

Department of Computer Science, Emory University, Atlanta, GA, USA · Department of Computer Science, Virginia Tech, Alexandria, VA, USA · Department of Electrical and Computer Engineering, Worcester Polytechnic Institute, Worcester, MA, USA+1

World Models Causal World Models

Open paper AskPDF arXiv

World models have emerged as a unifying paradigm for learning latent dynamics, simulating counterfactual futures, and supporting planning under uncertainty. In this paper, we argue that computational epidemiology is a natural and underdeveloped setting for world models. This is because epidemic decision-making requires reasoning about latent disease burden, imperfect and policy-dependent surveillance signals, and intervention effects are mediated by adaptive human behavior. We introduce a conceptual framework for epidemiological world models, formulating epidemics as controlled, partially observed dynamical systems in which (i) the true epidemic state is latent, (ii) observations are noisy and endogenous to policy, and (iii) interventions act as sequential actions whose effects propagate through behavioral and social feedback. We present three case studies that illustrate why explicit world modeling is necessary for policy-relevant reasoning: strategic misreporting in behavioral surveillance, systematic delays in time-lagged signals such as hospitalizations and deaths, and counterfactual intervention analysis where identical histories diverge under alternative action sequences.

Abstract

Corruption Robustness

Explore similar work

Toward World Models for Epidemiology

Integrating Persuasion Theory into the Epidemiological Modelling of Health Misinformation Spread on Social Media

Generating Public Health Responses using Survey-Augmented Large Language Models