Open-set face identification with index-of-max hashing by learning

Stable Hash Generation for Efficient Privacy-Preserving Face Identification

“…In contrast, the approach in [19] achieved competitive results with respect to its baseline (i.e., unprotected system) by employing different fusion strategies. Inspired by this idea, Sardar et al [21] introduced a novel hashing technique, namely FaceHashing.…”

“…According to Drozdowski et al [11], only a few works have combined WR strategies with BTP schemes to build large-scale identification systems that ensure privacy protection. In the context of face biometrics, those studies have mainly employed cancelable biometrics [19]- [21]. However, most of those systems still report a degradation w.r.t.…”

“…Recently, Dong et al [19] proposed an identification system where privacy was ensured by transforming facial features to more compact non-invertible features. As a consequence, faster comparison (1:1) in the Hamming space could be achieved.…”

“…It is worth noting that the aforementioned research efforts usually yield a trade-off between privacy protection, computational efficiency, and biometric performance within face identification systems. Some of these works still suffer from significant performance degradations (e.g., [19]). Other works do not provide sufficient privacy protection, i.e., they leak biometric feature information (e.g., [41], [42]).…”

“…In addition, not all methods have properly shown, that they can reach unlinkability for cross-comparison attacks [15]. Further, WR strategies employed by these systems (e.g., [19]- [21]) are limited to feature transformation or exhaustive pre-selection approaches (e.g., [25]).…”