Identifying the sites of disulfide bonds in a protein is essential for thorough understanding of a protein's tertiary and quaternary structures and its biological functions. Disulfide linked peptides are usually identified indirectly by labeling free sulfhydryl groups with alkylating agents, followed by chemical reduction and mass spectral comparison or by detecting the expected masses of disulfide linked peptides on mass scan level. However, these approaches for determination of disulfide bonds become ambiguous when the protein is highly bridged and modified. For accurate identification of disulfide linked peptides, we present here an algorithmic solution for the analysis of tandem mass (MS/MS) spectra of disulfide bonded peptides under nonreducing condition. A new algorithm called "DBond" analyzes disulfide linked peptides based on specific features of disulfide bonds. To determine disulfide linked sites, DBond takes into account fragmentation patterns of disulfide linked peptides in nucleoside diphosphate kinase (NDPK) as a model protein, considering fragment ions including cysteine, cysteine thioaldehyde (-2 Da, C(T)), cysteine persulfide (+32 Da, C(S)) and dehydroalanine (-34 Da, C(Delta)). Using this algorithm, we successfully identified about a dozen novel disulfide bonds in a hexa EF-hand calcium binding protein secretagogin and in a methionine sulfoxide reductase. We believe that DBond, taking into account the disulfide bond fragmentation characteristics and post-translational modifications, offers a novel approach for automatic identification of unknown disulfide bonds and their sites in proteins from MS/MS spectra.