AI Driven Predictive Cyber-Risk Scoring for IoT Deployments in Smart Cities (Machine Learning for real-time Risk Scoring)

Abstract

The development of Internet of Things (IoT) infrastructures in smart cities has already increased the magnitude, heterogeneity and interconnectedness of urban cyber-physical systems, thus exposing systems to systemic cyber-risk. Conventionally vulnerability-based scoring methods such as fixed severity methods are still inadequate to dynamically changing and large scale IoT ecosystems. The paper introduces a predictive, AI-based structure of cyber-risk scoring that will allow smart cities to scale its risk quantification towards infrastructure levels. The suggested framework combines predicting attack possibilities, situational effects, exposure, asset relevance, and network topology into a unified model of many factors’ aggregation. A time-based updating process is proposed to guarantee a controlled risk recalibration when the threat conditions change and a network-aware aggregation policy involves the cascading and systemic propagation of risks in the interconnected IoT infrastructures. In contrast to empirical research in the field of intrusion detection, the study is conceptual and analytical, allowing formalization in mathematics, and theoretical validation, based on scenarios, evaluation. Analytical outcomes reveal that there is a limited scope, proportional responsiveness, temporal consistency, and infrastructure-based distinction of risk representation. The framework also allows policy-controllable prioritization, without reference to particular implementation technologies or datasets. This research will transform current cyber-risk evaluation methods with reactionary metrics that assess individual devices to predictive systemic modeling of risk across networks in smart city existence. The suggested architecture also offers a systematic base to subsequent, practical application, elucidable AI fusion, and operational utilization in next-generation city structures.