In order to improve the identification level of hardware Trojan, a rare node activity improvement method based on genetic algorithm was proposed, by using the global searching ability of genetic algorithm. The proposed method was used to search the test vector sets which could improve the activity of rare nodes, by taking the toggle count of rare nodes as activity. Under the excitation of the test vector, the transition count of rare nodes in device under test (DUT) was taken as fitness, and the operations of selection, crossover and mutation were conducted during the whole test vector space. The fitness of parent and the fitness of offspring were compared, and the higher fitness test vector sets were saved, and eventually the optimized test vector set was generated by iteration when the iteration termination condition was reached. Simulation verification was conducted with the ISCAS'85 benchmark circuit c3540 as research object. Results showed that before the operation of the algorithm, when 1 000 vectors were used as input, the sum of the transition times of the un-rare nodes was 155 158, and the sum of the transition times of the rare nodes was 117. After the operation of the algorithm, when 1 000 vectors were used as input, the sum of the transition times of the un-rare nodes was 157 146, and the sum of the transition times of the rare nodes was 882. The test vectors generated by genetic algorithm improved the transition rate of rare nodes up to 7.54 times, and increased the relative transition rate of rare nodes up to 7.44 times.