Affected by the complexity of the power Internet of Things (IoT) and the stealth of terminal vulnerabilities, the traditional vulnerability correlation mining methods currently in use exhibit local biases in correlation feature parameters. This leads to insufficient overall mining scale and low global optimization efficiency of the algorithms, which severely impacts the normal operation of power IoT terminals. To address the aforementioned issues, starting from the structural characteristics of IoT, a black-box genetic algorithm is introduced. By completing the global parameter reconstruction and optimization of the overall mining method through four parts: power IoT terminal status perception, terminal vulnerability correlation mining rule generation, introduction of black-box genetic algorithm parameters, and terminal vulnerability correlation mining, the accuracy and scale of mining are enhanced. Simulation tests indicate that the mining curve values of the proposed method are relatively large, and the mean deviation index difference is 0.1. This demonstrates that the black-box genetic algorithm has high feasibility and effectiveness in the mining of security vulnerabilities in power IoT terminals, and the mining stability is sufficient to meet the current terminal vulnerability mining task requirements.