The invention provides an SQLite data recovery method applicable to merging of non-integral primary keys and free blocks and relates to the field of computer forensics. The method includes the steps of 1), starting from a root page and a file header to search all the free blocks and free pages; 2), analyzing difference of cell layout of integral primary keys and the non-integral primary keys as well as difference of meanings of bytes, temporarily assuming each free block contains one cell, calculating information of types covered, analyzing the type information and a data field and performing data recovery; 3), for the free blocks cannot pass the step 2), believing that each of the free blocks contains multiple cell, pre-estimating length of the cells stored in the free blocks is from 4 to free block length on the basis of splitting and verifying, and judging pre-estimated values with a single free block extraction module; 4), in response to decreasing of the free blocks due to B+tree reorganization, extracting the data from the free pages collected in the step 1). With the method, the data deleted can be easily recovered from SQLite3 database files, and the data is provided for follow-up electronic forensics work.