Challenges in constructing data sets for geographically informed infrastructure resilience analysis

Keywords: data handling, networks, infrastructure, hazards, risk, modeling
Abstract Type: Paper Abstract

Abstract

The US Cyberinfrastructure and Infrastructure Security Agency (CISA) is tasked with increasing the resilience of the nation's critical infrastructure, identifying and managing risks to lifeline systems (e.g., energy, communications, water) and critical functions of the nation from myriad human and natural hazards. Among the more challenging risks to model is cascading impact: interdependent infrastructure systems that sequentially fail because systems, commodities, or services they rely upon are no longer functional, and consequently downstream dependent systems are at risk. Modeling these relationships requires a cross-sectional and functional approach, including, crucially, data that contains up-to-date locational and attribute information for nodes of a network and links between these nodes, forming cross-sectional dependency chains. Unlike many geographic phenomena, this analytical realm is data sparse: collecting, verifying, and using relevant information relies on a mix of structured and unstructured data, developed for a multitude of purposes, by many different actors and analysts over many years. Handling these geographically referenced but heterogenous data sets is at the core of work that researchers at Idaho National Laboratory are performing for CISA. In our work, we leverage and deconflict existing data sets, examine multi-format sources to extract additional records and information, infer dependencies from engineering principles, expert-informed models, and spatial heuristics, and store the data in graph-style databases for use in geographic and logical maps and graphs. The challenges for creating a national, informative, geographic, and standardized data set of infrastructure (cyber and physical) will be presented, and INL's approaches for managing those challenges will be discussed.