Privacy-Aware Fingerprint Exposure Detection in Photographs Using Computer Vision and Vision APIs

Privacy-Aware Fingerprint Exposure Detection in Photographs Using Computer Vision and Vision APIs Mridul Banerjee UG student, Department of Electronics and Communication Engineering, Haldia Institute of Technology, Haldia, West Bengal ABSTRACT Photographs shared through social platforms can unintentionally expose biometric information when the friction ridge surface of a finger is visible. This work presents a lightweight prototype that identifies photographs with potential fingerprint exposure and applies a privacy-preserving image transformation before sharing. The proposed pipeline combines OpenCV for image loading, resizing, cropping and local masking; MediaPipe Hand Landmarker for locating hands and finger landmarks; and an optional vision API as a secondary semantic assessment layer. Instead of training a new deep-learning model, the system composes ready-made computer-vision components into a sense-assess-protect workflow. After a hand is localized, finger regions are estimated from the landmark geometry and relevant image crops are evaluated for visible biometric detail. When the risk exceeds a configurable threshold, the system automatically blurs or masks the sensitive region while leaving unrelated image content unchanged. The prototype is intended as an accessible demonstration of privacy-enhancing AI for cognitive systems rather than a certified biometric presentation-attack detector. Functional evaluation indicates that the integrated pipeline can detect hand-containing images, identify potentially exposed finger regions, and generate a protected output with limited modification to non-sensitive content. The study also highlights limitations caused by viewpoint, illumination, image resolution and the distinction between visible finger texture and a fingerprint image of matching quality.