The fuzzy vault scheme as an instance of a biometric cryptosystem is able to provide biometric authentication while also protecting privacy. In combination with multibiometrics the fuzzy vault scheme can potentially be used for applications that require a high security level. While many approaches for constructing multibiometric fuzzy vaults have been proposed there is a lack of a method that guarantees a balanced contribution of the individual characteristics to the fuzzy vault. Moreover, the potential security level of most approaches could only be extrapolated due to empirical testing being conducted with a limited amount of non-mated comparisons. This work aims to achieve three objectives. Firstly, craft a unified framework for multibiometric fuzzy vaults, that is able to construct balanced fuzzy vaults using any combination of characteristics. Secondly, investigate whether it is possible to empirically measure high security levels of multibiometric fuzzy vaults using a limited dataset. Thirdly, evaluate the proposed approach in terms of accuracy and security. The first objective was addressed by proposing a unified framework created through the utilization of a generalized approach for feature transformation, and employing deep feature extractors known for generating feature vectors of consistent representation. The second objective was addressed by proposing a method that improves the multibiometric evaluation for l characteristics exponentially increasing the number of non-mated comparisons with l through means of creating virtual subjects. Finally to fulfill the last objective an empirical evaluation using a multibiometric dataset composed of face, fingerprint and iris samples was conducted. To determine if the proposed framework and evaluation methods effectively achieved their intended objectives, further experiments were conducted investigating the effect of feature transformation and the impact of the improved evaluation on the measurable security level.