Numerical Study of the Dynamics of Medical Data Security in Information Systems

Background: Integrated medical information systems process large volumes of sensitive clinical data and are exposed to persistent cyber threats. Artificial intelligence (AI) is increasingly used for anomaly detection and incident response, yet its systemic effect on the dynamics of security indicators is not fully quantified. Aim: To develop and numerically study a nonlinear dynamical model describing the joint evolution of system vulnerability, threat activity, compromise level, AI detection quality, and response resources in a medical data protection context. Method: A five-dimensional system of ordinary differential equations was formulated for variables V, T, C, D, R. Parameters characterize appearance and elimination of vulnerabilities, attack intensity, AI learning and degradation, and resource consumption. The corresponding Cauchy problem V0=0.5, T0=0.6, C0=0.1, D0=0.4, R0=0.8 was solved on 0,200 numerically using a fourth-order Runge–Kutta method. Results: Numerical modelling showed convergence to a favourable steady regime. On the interval t ∈ [195, 200] the mean values were V=0.0073, T=0.3044, C=7.7·10−5, D=0.575, R=19.99. Thus, the initial 10% compromise is reduced by more than 99.9%, while AI detection quality stabilizes at around 0.58, and response capacity increases 25-fold. Conclusions: The model quantitatively confirms that the integration of AI detection and a managed response capacity enables the system to reach a stable state with virtually zero compromised medical data even with non-zero threat activity.