International Association
for Cryptologic Research

A noisy non-uniformly distributed secret often needs to be transformed into a stable high-entropy key. Biometric systems and physically unclonable functions (PUFs) exemplify the need for this conversion. Secure sketches are a useful tool hereby as they alleviate the noisiness while keeping the corresponding min-entropy loss to a minimum. The novelty of our work is twofold. First, seven secure sketch constructions, all based on a binary $[n,k,d]$ block code, are proven to be largely interchangeable. Despite having different looks and properties, all exhibit the same min-entropy loss, when fed with the same probability distribution. Second, for PUF-induced distributions with practical relevance, we derive new unified bounds on the min-entropy loss, considerably tighter than the more general well-known $(n-k)$ bound. Our bounds allow for an efficient evaluation and are hence suitable for reducing the implementation footprint of the sketch. This is beneficial for resource-constrained devices in particular.