Strong key derivation from noisy sources

A shared cryptographic key enables strong authentication. Candidate sources for creating such a shared key include biometrics and physically unclonable functions. However, these sources come with a substantial problem: noise in repeated readings. A fuzzy extractor produces a stable key from a noisy source. It consists of two stages. At enrollment time, the generate algorithm produces a key from an initial reading of the source. At authentication time, the reproduce algorithm takes a repeated but noisy reading of the source, yielding the same key when the two readings are close. For many sources of practical importance, traditional fuzzy extractors provide no meaningful security guarantee. This dissertation improves key derivation from noisy sources. These improvements stem from three observations about traditional fuzzy extractors. First, the only property of a source that standard fuzzy extractors use is the entropy in the original reading. We observe that additional structural information about the source can facilitate key derivation. Second, most fuzzy extractors work by first recovering the initial reading from the noisy reading (known as a secure sketch). This approach imposes harsh limitations on