"Has this person been encountered before?": Modeling an anonymous identification system

Brian DeCann, Arun Ross

Abstract: We consider the problem of anonymous identification where a biometric system answers the question “Has this person been encountered before?” without actually deducing the person's identity. In such a system, identity profiles are created dynamically as and when the system encounters an input probe. Consequently, multiple probes of the same identity may be mistakenly placed in different identity profiles, while probes from different identities may be mistakenly placed in the same identity profile. In this work, we model the matching performance of an anonymous identification system and develop terminology as well as expressions for predicting decision errors. Further, we demonstrate that the sequential order in which the probes are encountered by the system has a great impact on its matching performance. Experimental analysis based on face, fingerprint and iris scores confirms the validity of the designed error prediction model, as well as demonstrates that traditional metrics for biometric recognition fail to completely characterize the error dynamics of an anonymous identification system.