Discriminant analysis is an inappropriate method of authentication

• Authentication is considered as a methodological problem.
• Discriminant methods do a poor authentication job.
• They wrongly identify samples that do not belong to the predefined classes.
• Only one-class classifiers should be used for authentication.

Authentication is the process of determining whether an object is, in fact, what it is declared to be. In practice, this problem is often solved by using discriminant methods. In this paper, we explain that such techniques do a poor authentication job. The main drawback of discriminant methods is their inability of proper classification of new samples, which do not belong to any of the predefined classes. Our considerations are illustrated by a real-world example and a comparison of the results provided by the following two methods: Partial Least Squares- Discriminant Analysis, PLS-DA, and Data Driven Soft Independent Modeling of Class Analogy, DD-SIMCA.