Enhancing face recognition privacy through the integration of differential privacy and convolutional neural network

Muhammad Minoar Hossain, Mohammad Motiur Rahman · 2026

Read Enhancing face recognition privacy through the integration of differential privacy and convolutional neural network by Muhammad Minoar Hossain, Mohammad Motiur Rahman (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/plos-pone-0353565.txt.

Description

Muhammad Minoar Hossain and Mohammad Motiur Rahman design a face recognition system that cannot easily leak the faces it knows, in this 2026 paper. Databases of faces face real threats, from deepfakes to unauthorized access, and most recognition research ignores the privacy side. Their approach deliberately perturbs stored face images using differential privacy techniques, a mathematical way of adding protective noise, and trains a neural network built to learn from the scrambled images.

They tested three perturbation styles across a wide range of privacy strengths on two face datasets, pairing the network with several simple classifiers. A blur-based method with logistic regression gave the best balance, reaching high accuracy on the easier dataset while keeping the stored images obscured. The authors argue recognition and privacy need not be opposites if the network is designed for noisy inputs from the start.

Catalog record

CREATORMuhammad Minoar Hossain, Mohammad Motiur Rahman
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE38820
LICENSECC_BY
RIGHTSCC BY
SOURCEPLOS

Source: https://doi.org/10.1371/journal.pone.0353565

Browse the full library · About the Tower of Babel