Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{189521,
        author = {RAJESH S and Roland A and Sachithananthan S},
        title = {Identifying and Evaluating Soft-Biometric Privacy-Enhancement Methods},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {7},
        pages = {8032-8042},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=189521},
        abstract = {Protecting attributes, such as gender, in facial image is critical for enhancing privacy in “face recognition systems”, yet the resilience of privacy-preserving methods against attribute recovery remains insufficiently explored. This work presents Privacy-Prober, an innovative framework to evaluate and detect soft-biometric privacy enhancement techniques under black-box conditions. We employ the Fast Gradient Sign Method (FGSM) to suppress gender attributes and introduce novel recovery approach Averaging Multiple Adversarial Perturbations. Using the LFW, MUCT and Aidence dataset, we assess performance through Suppression Rate (SR), Identity Loss (IL), Privacy-Gain Identity-Loss Coefficient (PIC), Attribute-Recovery Robustness (ARR), and a simplified APEND detection mechanism. Results show strong SR (0.89–0.92) and favorable PIC (0.66–0.77), with Averaging surpassing denoising in privacy-utility balance. Privacy-Prober offers practical, lightweight tools for privacy assessment, highlighting FGSM vulnerabilities and guiding future advancements in secure biometric systems.},
        keywords = {Privacy Enhancement, Attribute Recovery, FGSM (Fast Gradient Sign Method), k-AAP, Averaging Recovery, Denoising Recovery, Inpainting Recovery.},
        month = {January},
        }