This paper considers the problem of robustly identifying m intruders in a network consisting of n cooperative agents which are subject to unknown disturbances. First, a distributed system model is introduced so that the relationship between agents, the attacks and unknown disturbances can be captured. Next, the distributed identification scheme is formulated as a spectral assignment problem and necessary filter gains are obtained through a carefully constructed linear system of equations. Necessary and sufficient conditions to decouple the unknown disturbances from the agents residual generators are derived in terms of filter gain matrices. It is shown that the problem of discriminating between unknown disturbances and attacks in a distributed system under consensus dynamics can be reduced to the problem of determining a set of constraints on the spectrum of the residual generator coefficient matrices. The approach is illustrated through an example.