Reliability assessment is an important way to evaluate the reliability of system, and it has been an essential way to control the reliability of system quantitatively. But there are some difficulties in achieving an objective result on reliability of complex system using the traditional way of statistics based on large samples due to time and cost. Thus, the reliability assessment of this kind of system has been a problem in reality. Considering the problem met in the reliability assessment and the life prediction for complex system, the reliability information fusion method based on mean interactive entropy is proposed. Based on mean interactive entropy, the weight of fusion was achieved, and the reliability of system was assessed by fusion of multiple prior distributions. By reducing uncertainty in the process, the new method increased the dependence of assessment and was less subjective. Finally, a simulation example was presented to illustrate the efficiency of the fusion method.