Exchange of chemical information without disclosure of the respective structures would greatly increase the data sets available for model building. Within the framework of the ChemMask project we explored the principal applicability of SIBAR-descriptors to mask chemical structures. SIBAR is based on calculation of similarity values for each compound of the training set to a set of reference compounds. Although the SIBAR-approach per se does not allow to unambiguously trace back the chemical structure of a compound, similarity searching in a 1.5 million compound database spiked with compounds structurally analogous to the query structure lead to the retrieval of compounds structurally and pharmacologically highly analogous to the "hidden" query structure in all three examples investigated. Comparison to results obtained with the original descriptors used to calculate the SIBAR-values showed, that SIBAR indeed adds some fuzziness to the data matrix.