In this paper, we first define bisimulation-based non-deterministic admissible interference (BNAI), derive its process-theoretic characterisation and present a compositional verification method with respect to the main operators over communicating processes, generalising in this way the similar trace-based results obtained [J. Univ. Comput. Sci. 6 (2000) 1054] into the finer notion of observation-based bisimulation [Logic and Models of Concurrent Systems, 1985]. Like its trace-based version, BNAI admits information flow between secrecy levels only through a downgrader (e.g. a cryptosystem), but is phrased into a generalisation of observational equivalence [Communication and Concurrency, 1989]. We then describe an admissible interference-based method for the analysis of cryptographic protocols, extending, in a non-trivial way, the non-interference-based approach presented by Focardi et al. [Proceedings of DERA/RHUL Workshop on Secure Architectures and Information Flow, 2000]. Confidentiality and authentication for cryptoprotocols are defined in terms of BNAI and their respective bisimulation-based proof methods are derived. Finally, as a significant illustration of the method, we consider simple case studies: the paradigmatic examples of the Wide Mouthed Frog protocol [ACM Trans. Comput. Syst. 8 (1990) 18] and the Woo and Lam one-way authentication protocol [IEEE Comput. 25 (1992) 39]. The original idea of this methodology is to prove that the intruder may interfere with the protocol only through selected channels considered as admissible when leading to harmless interference.