Adaptive Runtime Risk Assessment With Aging-Aware Feasibility for Safety-Critical Systems

In safety-critical systems, such as vehicles, continuous and adaptive risk assessment is essential for maintaining security throughout the system lifecycle. Traditional Threat Analysis and Risk Assessments (TARAs), confined to the design phase of a system, fail to address unforeseen vulnerabilities during operation. This paper introduces a framework for continuous runtime risk assessment, aligned with the automotive standard ISO/SAE 21434 as a concrete application scenario, combining classical TARA during initialization and an adapted version during runtime. The runtime phase updates a preweighted Attack Tree (AT) based on emerging vulnerabilities and applies an aging function to adjust risk feasibility using the Common Vulnerability Scoring System (CVSS). The novelty of the framework lies in its runtime adaptation capability to new threats. We evaluate our framework using representative automotive architectures and attack scenarios and further conduct an extensive evaluation by applying it to a generic autonomous vehicular architecture. Results demonstrate that the system delivers timely and adaptive risk assessments, effectively manages evolving cybersecurity threats, and offers a resource-efficient solution suitable for real-world deployment.