Abstract: A method based on an improved artificial immune strategy is introduced for the optimization of distributed multi-sensor decision fusion systems under Neyman-Pearson criteria for the cases with statistically dependent observation and fixed fusion rule. The object function is optimized in two steps without any information of its derivation:filter operator is used for pre-search to reduce the search space and then an artificial immune strategy is applied for the global search. The experimental results show that the proposed method has better convergence and higher precision than the traditional gradient algorithms. A further discussion on the best fusion rule for different means of signals is given.