计算机应用 ›› 2014, Vol. 34 ›› Issue (5): 1383-1385.DOI: 10.11772/j.issn.1001-9081.2014.05.1383

• 人工智能 • 上一篇    下一篇

谱聚类递归神经网络集成的全自动公开区分计算机和人的图灵测试识别算法

张亮,陈睿,邱小松   

  1. 电子工程学院 网络系,合肥 230037
  • 收稿日期:2013-11-01 修回日期:2013-12-17 出版日期:2014-05-01 发布日期:2014-05-30
  • 通讯作者: 张亮
  • 作者简介:张亮(1982-),男,湖南醴陵人,讲师,博士,主要研究方向:CAPTCHA识别、信息安全;陈睿(1986-),男,江西抚州人,博士研究生,主要研究方向:CAPTCHA识别、信息安全;邱小松(1987-),男,湖北随州人,硕士研究生,主要研究方向:协议分析、信息安全。
  • 基金资助:

    安徽省自然科学基金资助项目

Completely automated public turing test to tell computers and humans apart recognition algorithm based on spectral-clustering recurrent neural network ensemble

ZHANG Liang,CHEN Rui,QIU Xiaosong   

  1. Department of Network, Electronic Engineering Institute, Hefei Anhui 230037, China
  • Received:2013-11-01 Revised:2013-12-17 Online:2014-05-01 Published:2014-05-30
  • Contact: ZHANG Liang

摘要:

针对粘着全自动公开的区分计算机和人的图灵测试(CAPTCHA)的识别问题,提出了一种基于谱聚类递归神经网络(RNN)集成的识别算法。该算法首先使用不一致测度度量两个RNN之间的距离,构建出一张由多个候选RNN形成的图;然后基于谱图聚类理论,将多个RNN划分为不同的簇,并在每个簇上选择最佳RNN参与集成。实验结果表明:相对于单个候选RNN,该算法的识别率提高了约16%;相对于全部候选RNN构成的集成系统,该算法形成的集成规模更小,仅为原来的23%。

Abstract:

Concerning the recognition of closely-connected Completely Automated Public Turing Test to Tell Computers and Humans Apart (CAPTCHA), a recognition algorithm based on spectral-clustering Recurrent Neural Network (RNN) ensemble was proposed. This algorithm firstly used disagreement measure for distance between two RNNs, thus constructed a graph composed by candidate RNNs. Then, a graph cluster method was used to divide RNNs into clusters. Finally, the best RNN in each cluster was selected. The experimental results reveal that: compared with single candidate RNN, recognition rates of this algorithm is increased by 16%. Compared with the ensemble of all candidate RNNs, ensemble size of this algorithm is much smaller, it is about 23% of the original size.

中图分类号: