We used the MCS paradigm in the following application fields:

  • Biometric identity recognition: we developed novel algorithms and frameworks for combining information coming from different biometric traits, in multi-modal biometric systems (mainly based on faces and fingerprints, but also on soft biometrics like hair colour); we also developed adaptive multi-biometric systems based on template co-update techniques.
  • Computer and network security: we developed network-based intrusion detection systems capable of reducing the number of false positives and improving the detection rate, by exploiting the diversity of the individual classifiers of a MCS. We also developed an advanced intrusion detection system for web services, made up of an ensemble of specific anomaly detectors: such an MCS architecture allows us to reduce the complexity of the detection problem by separately modelling each category of anomalies, and provides a human-readable representation of anomalous events.
  • Content-based image retrieval: we developed relevance feedback techniques based on combining different feature sets (feature-level fusion).
  • Classification of text, documents and images: we exploited MCSs for developing multi-label classification techniques based on classifier selection.