Implementation and Analysis of Homomorphic Facial Image Encryption and Manipulation
Homomorphic cryptography allows for encrypted data to be operated on securely without requiring decryption. Current homomorphic cryptosystems are either limited in their permitted operations or are significantly more time-intensive than standard non-homomorphic cryptosystems. Regardless, homomorphic cryptography has been targeted for use in secure image processing and facial recognition, due to their ability to maintain data privacy. In this paper, we assessed the viability of the Paillier, Damgård--Geisler--Krøigaard (DGK), and Brakerski--Gentry--Vaikuntanathan (BGV) cryptosystems for facial image processing applications, by implementing a software library equipped with these cryptosystems, and comparing their time efficiency and accuracy. As an extension to previous research, we attempt to support non-linear image processing operations. Preliminary results have shown that the Paillier and DGK cryptosystems are comparable in accuracy and may be used for image negation, but only the Paillier cryptosystem is consistent enough to produce reasonable to accurate results.
Asuncion, A. E. C., Guadalupe, B. C. T., & Yu, W. E. S. (2019). Implementation and analysis of homomorphic facial image encryption and manipulation. Proceedings of the 2019 4th International Conference on Multimedia Systems and Signal Processing, 158–166. https://doi.org/10.1145/3330393.3330407