Researchers say scheme improves homomorphic encryption efficiency, effectiveness

Homomorphic encryption has held elusive promise for years, and recent research and developments continue to suggest that a practical version of the technology is not far off.

A new paper on “Efficient face information encryption and verification scheme based on full homomorphic encryption” published in Scientific Reports proposes a way to address the efficiency, accuracy and security challenges of full homomorphic encryption with face biometrics. The paper was written by a team of researchers from Changchun University of Science and Technology and the Huayin Ordnance Test Center in China.

The researchers identify three barriers to compliance remaining for homomorphic encryption, starting with the rapid expansion of cyphertext beyond practical volumes in high-dimensional facial features. They note an 83 percent failure rate in 2024 audits documented by ENISA. And security mechanisms will also have to be developed to protect against IND-CPA (indistinguishability under chosen-plaintext attack) and IND CCA2 (indistinguishability under adaptive chosen ciphertext attack) attacks in line with the NIST SP 800-57 standard for key management guidelines and ISO 19772 for authenticated encryption.

To address these problems, they propose a Hybrid Encryption with Facial Data Integrity Verification Scheme (HEFDIVS). The scheme applies dimensionality reduction, which contributes to an 89 percent improvement in ciphertext computation time. It also makes use of hybrid encryption algorithms combining SM2 and SM4 to protect against the sophisticated attacks noted above.

The resulting system reaches mean accuracy rates of 95.45 percent and 96.98 percent on the LFW and Faces94 biometric datasets, respectively, according to the paper.

Homomorphic encryption is applied in the market to proof of humanity by Privasea, and it shows up in an IEEE standard for biometric template security, but the technology’s sceptics seems to outnumber its customers so far.

Article Topics

biometrics  |  biometrics research  |  data protection  |  face biometrics  |  homomorphic encryption