Security for Distributed Deep Neural Networks in Miniature Fingerprint Recognition Utilizing a Genetic Algorithm

The most traditional and popular biometric identification method is based on fingerprints. Everybody has a set of unchanging, distinctive fingerprints. It is essential to label minutiae appropriately and reject fake ones since recognition systems rely on local ridge characteristics. As a result, fingerprint pictures need to be controlled for clarity, minute results need to be computed, and then templates need to be compared to templates that are kept in the database. The research suggested three methods to enhance images, extract information, and match with fingerprint templates. The study’s first phase included 300 x 300 neural network picture inputs with various fingerprints gathered from two sets of false and actual images. In order to anticipate the optimal miniature extraction using genetically based chromosomal matching miniature analysis, we then use the VGG16 model. The suggested Model predicts 96.20 for smaller datasets, and when the size of the dataset was enlarged, the accuracy decreased to 86.89 percent, which may be employed in practical applications and contributes to system security.