Biometric and Physical Identifiers with Correlated Noise for\n Controllable Private Authentication

The problem of secret-key based authentication under privacy and storage\nconstraints on the source sequence is considered. The identifier measurement\nchannels during authentication are assumed to be controllable via a\ncost-constrained action sequence. Single-letter inner and outer bounds for the\nkey-leakage-storage-cost regions are derived for a generalization of a classic\ntwo-terminal key agreement model with an eavesdropper that observes a sequence\nthat is correlated with the sequences observed by the legitimate terminals. The\nadditions to the model are that the encoder observes a noisy version of a\nremote source, and the noisy output and the remote source output together with\nan action sequence are given as inputs to the measurement channel at the\ndecoder. Thus, correlation is introduced between the noise components on the\nencoder and decoder measurements. The model with a secret key generated by an\nencoder is extended to the randomized models, where a secret-key is embedded to\nthe encoder. The results are relevant for several user and device\nauthentication scenarios including physical and biometric identifiers with\nmultiple measurements that provide diversity and multiplexing gains. To\nillustrate the behavior of the rate region, achievable (secret-key rate,\nstorage-rate, cost) tuples are given for binary identifiers and measurement\nchannels that can be represented as a mixture of binary symmetric subchannels.\nThe gains from using an action sequence such as a large secret-key rate at a\nsignificantly small hardware cost, are illustrated to motivate the use of\nlow-complexity transform-coding algorithms with cost-constrained actions.\n