Multimedia Tools and Applications https://doi.org/10.1007/s11042-023-16511-6 CLNet: a contactless fingerprint spoof detection using deep neural networks with a transfer learning approach Kanchana Rajaram 1 · Bhuvaneswari Amma N.G. 2 · Ashwin S. Guptha 2 · Selvakumar S. 3,4 Received: 11 August 2022 / Revised: 7 June 2023 / Accepted: 9 August 2023 © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2023 Abstract Biometric fingerprint verification and identification have been extensively used in real life applications as an authentication and access control mechanism. Newer contactless finger- print scanning technology offers high convenience and hygiene, especially in the view of COVID-19. Attackers still challenge the biometric security offered by contactless scan- ners by illegitimate acquisition of the user’s fingerprint through various spoofing methods. Therefore, detection of contactless fingerprint spoof is on the urge to protect the biomet- ric security systems. The existing solutions to contactless fingerint spoof detection face the lacuna of considering limited fingerprint features leading to low spoof detection accuracy. In this study, this issue has been addressed and CLNet (Contact Less Network) approach is proposed to detect the spoofness in contactless fingerprints. The proposed CLNet is a deep neural network approach utilizing contactless fingerprint images followed by a transfer learning approach called SpoofDetNet which is based on the MobileNetV2 model. The moti- vation for the development of the SpoofDetNet is to create a spoof detection method viable for contactless fingerprint images as well as contact-based fingerprint images which stand strong among state-of-the-art models. We created a Spoofed-Contactless Adult Fingerprint (S-CLAF) dataset with live and spoof contactless fingerprint images. The CLNet approach was trained and tested on S-CLAF dataset and it achieved an accuracy of 99.07% across all spoofed materials. Furthermore, the proposed approach was tested using LivDet 2015 bench- mark dataset and IIT Bombay touchless fingerprint dataset achieving accuracy of 98.32% and 99.38% respectively. It is evident from the experimental results that the proposed CLNet outperforms the state-of-the-art fingerprint spoof detection methods. Keywords Biometric security · Contactless fingerprints · Convolutional neural networks · Fingerprint spoof detection · Transfer learning B Kanchana Rajaram rkanch@ssn.edu.in Extended author information available on the last page of the article 123