Singularly perturbed systems are the mathematical models of many technical processes. The available various asymptotic schemes of the construction of suboptimal control laws for such the systems can not be frequently used in incomplete assumed information of process running. The problem of minimax filtering is considered for singularly perturbed linear systems with variable coefficients operating in the presence of uncertain actions in the integral quadratic constraints on realization of the lasts. The asymptotic estimates of the set of possible system states compatible with the observed signal are calculated.