POSH : a generalized CAPTCHA with security applications

Abstract

A puzzle only solvable by humans, or POSH, is a prompt or question with three important properties: it can be generated by a computer, it can be answered consistently by a human, and a human answer cannot be efficiently predicted by a computer. In fact, a POSH does not necessarily have to be verifiable by a computer at all. One application of POSHes is a scheme proposed by Canetti et al. that limits on-line dictionary attacks against password-protected local storage, without the use of any secure hardware or secret storage. We explore the area of POSHes, implement several candidate POSHes and have users solve them, to evaluate their effectiveness. Given these data, we then implement the above scheme as an extension to the Mozilla Firefox web browser, where it is used to protect user certificates and saved passwords. In the course of doing so, we also define certain aspects of the threat model for our implementation (and the scheme) more precisely.