Bayesian Optimization with SAT Solver for Enhanced Database Security Identification

Abstract

To solve the limitations of traditional database security detection, such as weak logical reasoning, delayed feature update, and high computational complexity under high-dimensional constraints, this study proposes a Bayesian optimization algorithm based on SAT. Security rules are formalized into CNF logical constraints, enabling automated reasoning and rule consistency verification via a satisfiability solver. A Gaussian process-driven Bayesian optimization framework with an improved Expected Improvement (EI) acquisition function dynamically updates feature weights and accelerates convergence toward high-risk regions, enhancing rare vulnerability detection and global solving efficiency. Experiments on enterprise database logs and national vulnerability datasets demonstrate an average detection rate of 97.5%, a defense success rate of 95.8%, and a response latency of 2.11 s, outperforming baseline methods. The system stability index reaches 0.93, with concurrent processing capability improved by 34.2%. These results confirm the algorithm’s high-precision identification, stable defense performance, and practical deployability for database security management under complex, high-dimensional attack scenarios.