ANB-PW: An Adaptive Distinguishing Attack Method based on Naive Bayes

Abstract

Honeywords are decoy passwords placed in user’s account to detect whether password files are leaked. The distinguishing attack of honeywords plays a very important role in evaluating the strength of honeywords. However, the existing attack methods cannot simulate the ability of the attacker, leading to inaccurate evaluation of honeywords strength. In this paper, we propose an adaptive distinguishing attack method ANB-PW to construct a Naive Bayes classifier based on passwords and honeywords. In addition, ANB-PW uses guessed passwords to adjust its guessing strategy in real time. Our results suggest that compared with Top-PW and PCFG, ANB-PW has better attack ability.