Aiming at the problem that the seed selection of coverage-guided fuzzing test cannot reflect the value of seed mutation, a fitness function calculation method based on the potential of seed mutation was proposed. By assigning higher weights to the basic blocks which is close to the starting block of the program or has more subsequent blocks and tracking basic block information that is not covered near the coverage path of the seed,the seed fitness was calculated according to the weight of uncovered basic blocks and the seeds with high fitness and low resource overhead were selected for the next generation of mutation. The proposed fuzzing technique was compared with AFL on LAVA-M dataset and real Linux program. The results show that the code coverage, speed of vulnerability discovery and vulnerability discovery of this method are significantly improved while reducing resource overhead.Thus proves the effectiveness of the above screening strategy.