Resilient Data : An Interdisciplinary Approach

Cybersecurity continues the migration toward data-informed solutions and the quality of the data is gaining in importance. Data accuracy is foundational to the trustworthiness required of artificial intelligence solutions. Trustworthy data must be accurate, robust, resistant and resilient to unauthorized modifications, going beyond traditional security solutions that perform data integrity checking. Cyber-physical systems present unique challenges with physical outcomes.Cyber-physical systems present unique challenges in achieving trustworthy data. The combination of security data and safety data is unique. Capturing both sets of data and determining the accuracy of that data requires an interdisciplinary approach. This effort describes the merging of information theory and information security constructs along with physical systems data. Knowledge is needed in control systems engineering, cybersecurity and information theory. As training data that informs decisions and feeds artificial intelligence algorithms accuracy and resilience are important.Resilient data is trustworthy data that represents a research challenge offering an opportunity to apply lessons learned from information disorders into the broader cybersecurity environment including the cyber-physical systems that power much of the US critical infrastructure. Creating resilient data, for use as training data requires data be examined in ways that have not historically been a part of traditional cybersecurity analysis. This effort describes a proposed method of contextually evaluating cyber-physical systems security data in order to determine the accuracy of the data and in the event of tampering, reconstitute the data to the last known trustworthy state.