Automated Biometric Identification using Dorsal Hand Images and Convolutional Neural Networks

Abstract The identification of perpetrators, present in Child Sexual Abuse Imagery (CSAI), is a significant challenge due to the use anonymisation techniques that mask their identities. Consequently, researchers have investigated uncommon biometric identifiers such as knuckle patterns, palmprints and dorsal side hand. This research proposes Convolutional Neural Network (CNN) based, fully automated approach using hand images. performance three different CNN architectures, AlexNet, ResNet50 ResNet152, experimentally determined against two similar datasets, 11k Hands IITD databases. A transfer learning used final output layers CNNs are modified match number classes datasets. results showed ResNet achieved accuracies greater than 99.9% on both whereas AlexNet between 80.1% 93.7%. These demonstrate it feasible deep, off-the-shelf CNNs, ResNets, for highlights potential images identify perpetrators child sexual abuse from CSAI.