Blind SQL Injection is a type of SQL injection, which is an extremely common attack method on the Internet today and can lead to security issues such as data leakage and command execution. The efficiency of blind SQL injection is severely restricted due to its large number of requests and request traffic. The more advanced methods in this field include the binary search and the language model method. In order to solve the shortcomings of the binary search that ignores the logical relationship between characters and the language model method causing the request payload to increase, a dynamic selection algorithm is designed to make full use of the advantages of binary search and language model methods for different injection objects, and the search load is shortened by modifying the Huffman search algorithm. In terms of test performance, this algorithm reduces the average number of requests by about 19.7% compared to binary search; in terms of unordered string prediction, this algorithm reduces the average request length by 20.2% compared to the Hakuin algorithm.