Machine Learning and Pattern Recognition Fingerprint Spoofing 1 Introduction 1.1 Abstract This report aims to examine a dataset that contains various low-level images of real (defined as Au- thentic) and fake (defined as Spoofed) fingerprints using different Machine Learning models. First, we will explore the structure of the dataset and try to understand how it is distributed. Then, we will evaluate various classifiers and try to find the best system that can accurately classify our samples with the lowest cost. 1.2 First considerations about the problem The dataset is composed of samples that represent fingerprint images through low-dimensional repre- sentations called embeddings. Each fingerprint is represented by a 10-dimensional vector of continuous

Fig: 1