AI-Driven situational intelligence for national security: an adaptive defensive information system with multi-source data fusion and predictive analytics

Abstract

Existing defensive information systems for national security often suffer from high latency, domain-specific limitations, and limited adaptability to evolving threats (static cyber intrusion detection or single-domain monitoring). This study proposes a novel AI-driven situational intelligence model that integrates multi-source data fusion, predictive analytics, and reinforcement learning to address these gaps. Using a design science research (DSR) methodology, the model was conceptualized, implemented, and evaluated through simulated and classified datasets across four threat domains: cyber intrusion, UAV infiltration, maritime security breach, and disinformation campaigns. The architecture combines probabilistic data fusion, deep neural networks, ensemble classification, and LSTM-based forecasting, with reinforcement learning for adaptive response selection. The system achieved an average F1-score improvement of 4.3% over state-of-the-art baselines, reduced latency by up to 39%, and maintained >93% accuracy under degraded conditions (20% data loss, 500 Ms sensor delay, noise injection). An automated feedback loop improved performance on novel threats by 15.5% without manual retraining, demonstrating operational adaptability. The framework offers a low-latency, cross-domain, and adaptive intelligence capability for proactive threat mitigation. Its integration of predictive modeling with autonomous response protocols supports early threat neutralization and aligns with contemporary defense doctrines. This research introduces a novel, operationally deployable architecture that uniquely combines dynamic multi-source data fusion, real-time predictive analytics, and adaptive reinforcement learning in a unified defensive information system, advancing both the technical frontier and real-world applicability in AI-driven national defense.