A novel approach named recursive evidence filtering for information processing in distributed sensor networks (DSNs) is presented. It is based on the Dempster-Shafer (DS) formalism, facilitating evidence exchange and fusion in a simple and straightforward manner, while accommodating for partial and incomplete information. Certain restrictions on the coefficients impose several challenges in the design of such evidence filters suggesting that arbitrary frequency shaping is not possible. A general design procedure for IIR type evidence filters is presented with examples