A Novel Approach to Proactive Password Checking

Abstract

In this paper we propose a novel approach to strength password-based access control strategies. We describe a proactive password checker which uses a perceptron to decide whether a user’s password is easy-to-guess. The checker is simple and efficient, and it works since easy and hard-to-guess passwords seem to be linearly separable. Experimental results show that the error rates in many cases are close to zero, memory requirements can be quantified in few bytes, and the answers to classification queries are almost immediate. This research opens new directions to investigate on the applicability of neural network techniques to data security environments.