Recently, multi-source information fusion architecture has become a trend in the design of complex systems as the data processing technology developed rapidly. Although information fusion can bring many earnings, the system also suffers many dangers caused by false and possibility deliberately deceptive information. Meanwhile, the work environment tends to be complicated. The reliability of each basic information source may be fluctuant in the task process, which can directly influence the system risk. Therefore, it is necessary to make a trade-off between expected yield and design loss by considering the fluctuant reliability of basic information sources. Oriented for multi-source information system, this paper proposed a risk analysis and trade-off design method. In the method, considering the fluctuant reliability of basic information sources, expected yield is evaluated based on Dempster-Shafer evidence theory and then balanced with design loss. Finally, two numerical examples are presented. The question raised by this study is significant and the proposed methods could guide the optimization process for the system design.