Detecting Cybersecurity Risks Using Fuzzy Features and Incomplete Signatures of Computer Attacks Against Smart City Artificial Intelligence

Abstract

This paper integrates neural networks with fuzzy set theory - specifically fuzzy relation intersections - to address a two-phase cybersecurity risk detection problem characterized by ambiguous features and incomplete attack signature databases in Smart City artificial intelligence. This approach progressively eliminates and mitigates ambiguity (fuzziness), incompleteness, and inconsistency in the initial data required to rapidly and reliably detect and identify risks. The first phase processes fuzzy data regarding features of computer attacks on artificial intelligence, while the second phase handles incomplete or inconsistent information about existing and matching attack signatures in the database. These findings will help identify the presence and severity of specific attack features, thereby enhancing the reliability of feature detection and signature generation for both existing and prospective cyberattacks on Smart City artificial intelligence systems. Furthermore, we outline the results of experimental calculations, which validate the practical applicability of the proposed decision-making mechanisms for detecting cybersecurity risks.