How much assurance does a PIN provide?

  • Authors:
  • Jon Bentley;Colin Mallows

  • Affiliations:
  • Avaya Labs, Basking Ridge, NJ;Avaya Labs, Basking Ridge, NJ

  • Venue:
  • HIP'05 Proceedings of the Second international conference on Human Interactive Proofs
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

We would like to quantify the assurance contained in an authentication secret. For instance, how much assurance does a customer convey to a bank by revealing that his Personal Identification Number (PIN) is 1111? We review a number of previously proposed measures, such as Shannon Entropy and min-entropy. Although each is appropriate under some assumptions, none is robust regarding the attacker’s knowledge about a nonuniform distribution. We therefore offer new measures that are more robust and useful. Uniform distributions are easy to analyze, but are rare in human memory; we therefore investigate ways to ”groom” nonuniform distributions to be uniform. We describe experiments that apply the techniques to highly nonuniform distributions, such as English names.