Abstract: Fault diagnosis agreement (FDA) can maintain the performance and integrity of highly reliable distributed systems. However, most of previous FDA protocols only take into account simple network with single faulty component. It is more important to study complicated network with faulty nodes and faulty links for real distributed applications. Unfortunately, the diagnosis of malicious (Byzantine) fault components can not satisfy FDA in this situation because of the arbitrariness of its behavior. Thus, the model of invalid link is proposed firstly in this paper, which can more accurately describe the effect of malicious faulty component under network with dual faulty components, and improve fault diagnosis coverage. Afterwards, based on the invalid link model, an evidence-based fault diagnosis protocol, PLFDA, is presented. PLFDA collects the messages which have accumulated in a Byzantine agreement protocol as evidence and then diagnoses the set of faulty components by examining the collected evidences. It can not only detect and locate simultaneously both faulty nodes and faulty links, but also satisfy requirements of FDA in a synchronous fully connected network, where the number of allowable faulty components is not greater than n?2-1, of which the number of allowable faulty nodes is less than or equal to (n-1)?3. In addition, the proof of correctness and complexity of PLFDA and experimental results are given in the end.