Abstract: Considering the fault detection rate (FDR), faults isolation rate (FIR) and false alarm rate (FAR),this article presents a universal mathematical model to solve test selection problem under unreliable test based on diagnostic dictionary matrix. On the basis of the contribution of FDR and FIR, a new function named the fault detection and isolation fee (FDIF) was defined and it was adopted as the greedy criterion. A new dynamic greedy algorithm was proposed to deal with the test selection issue with unreliable test. An application case was given to verify the algorithm. It shows that the algorithm can find out the optimal or suboptimal test set and it is appropriate to analysis or verification of the testability in large complex systems.