• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文学位 >Assessing and reducing spoofing vulnerability for multimodal and fingerprint biometrics.
【24h】

Assessing and reducing spoofing vulnerability for multimodal and fingerprint biometrics.

机译：评估和减少多模式和指纹生物识别技术的欺骗漏洞。

获取原文
获取原文并翻译 | 示例
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Biometric systems utilizing personal unique physiological or behavioral characteristics have become widely popular in providing security for information technology and entry to sensitive locations like airports, finance, health care, government, military, and any other type of business that are connected by a network. However, biometric recognition degrades in the presence of a mismatch between training and testing conditions. For face recognition, it can be in the form of pose change, illumination and expression changes. For speech recognition, that is usually in the form of channel distortion, environmental noise, health or emotional changes. In addition, these systems have been proven to be vulnerable to spoofing attacks, or granting entry to an impostor. Typically, an impostor is defined as a person who is not the user of the biometric system. Here, we also evaluate imposters deliberately attempting to gain access through presentation of a high resolution face photo, a voice recording tape or a fake finger. Liveness detection in a biometric system ensures that only "real" fingerprints, facial images, voice, iris, and other patterned characteristics are capable of generating templates for enrollment, verification, and identification. The objective of this research is to investigate methods to defeat this kind of spoof or impostor attacks for (1) a face/voice system and (2) a fingerprint system.;In the first part, the possibility of using lip motion as a biometric trait for a face/voice system is investigated. Lip motion can be integrated with the current audio-visual biometric system using face and voice recognition to build a multimodal biometric system that can improve recognition in some non-ideal conditions. In addition, using lip motion is a natural liveness testing solution to defeat some simple spoof attacks, like still face photo and recorded voice tape. In the case of the impostor using recorded lip motion video, we propose to utilize speech recognition as a challenge test, ensuring that the system would only allow the subject to pass when both multimodal biometrics recognition is successful and the subject speaks the randomly prompted password correctly.;Multimodal biometrics have the advantages of improving matching accuracy, addressing the issue of non-universality or insufficient population coverage, addressing the problem of noisy data, and providing anti-spoofing protection for the biometric system. However, no one has done any tests or evaluation on the spoofing vulnerability of a multimodal system. In this thesis, multimodal fusion of lip motion, voice and face is implemented for different kinds of spoof scenarios including traditional impostor attacks, and one spoof or two spoof attempts. A strategy for assessing spoofing vulnerability in a multimodal system is proposed considering that one or partial biometrics are spoofed. Non-zero effort (spoof) false acceptance rate is suggested as a evaluation metric for a multimodal biometric system. To reduce spoofing vulnerability, an appropriate operating point needs to be optimized in case of one or two biometrics are spoofed.;In the second part, the algorithms to defeat spoof attacks for fingerprint biometrics are investigated. One highly publicized vulnerability is that it is possible to spoof a variety of fingerprint scanners using artificial fingers made from Play-Doh, gelatin and silicone molds. Liveness detection is one type of anti-spoofing method, which measures physiological signs of life from fingerprints ensuring only live fingers are captured for enrollment or authentication. In this part, several novel new liveness detection methods have been proposed.;In summary, this work makes 3 main contributions. (1) A framework for assessing spoofing vulnerability in multimodal biometric system fusion of lip/voice/face signatures was developed. (2) A robust multimodal biometric system fusing face, voice and text-independent lip motion was developed. (3) Several novel liveness detection algorithms were developed based on valley noise, static perspiration pattern and intensity analysis. (Abstract shortened by UMI.)
机译：利用个人独特的生理或行为特征的生物识别系统在为信息技术提供安全性以及进入敏感位置（如机场，金融，医疗保健，政府，军事以及通过网络连接的任何其他类型的业务）方面已经广受欢迎。但是，在训练条件和测试条件不匹配的情况下，生物识别会降低。对于面部识别，可以采取姿势变化，照明和表情变化的形式。对于语音识别，通常采取通道失真，环境噪声，健康或情绪变化的形式。此外，这些系统已被证明容易受到欺骗攻击或允许冒充者进入。通常，冒名顶替者被定义为不是生物识别系统用户的人。在这里，我们还将评估冒名顶替者，他们试图通过呈现高分辨率的面部照片，录音带或假手指来获得访问权。生物识别系统中的活动检测可确保只有“真实的”指纹，面部图像，语音，虹膜和其他图案特征才能生成用于注册，验证和识别的模板。这项研究的目的是研究针对（1）脸部/语音系统和（2）指纹系统来克服这种欺骗或冒名顶替者攻击的方法。在第一部分中，使用嘴唇运动作为生物特征识别的可能性面部/语音系统的特征进行了研究。可以将唇部运动与使用面部和语音识别的当前视听生物识别系统集成在一起，以构建可以改善某些非理想条件下的识别的多模式生物识别系统。此外，使用唇部运动是一种自然的活力测试解决方案，可以克服一些简单的欺骗性攻击，例如静止图像和录音带。对于冒名顶替者使用录制的嘴唇运动视频的情况，我们建议将语音识别用作挑战测试，以确保系统仅在两种多模态生物特征识别均成功且受试者正确说出随机提示的密码时才允许受试者通过。;多模式生物识别技术的优点是提高匹配精度，解决非通用性或人口覆盖率不足的问题，解决嘈杂的数据问题以及为生物识别系统提供反欺骗保护。但是，没有人对多模式系统的欺骗漏洞进行任何测试或评估。本文针对传统的冒名顶替者攻击，一次或两次欺骗尝试，对不同形式的欺骗场景实现了嘴唇运动，声音和面部多模式融合。考虑到一个或部分生物特征被欺骗，提出了一种评估多模式系统中欺骗脆弱性的策略。非零努力（欺骗）错误接受率被建议作为多模式生物识别系统的评估指标。为了减少欺骗的脆弱性，在欺骗一个或两个生物特征的情况下，需要优化一个合适的工作点。第二部分，研究了克服指纹生物特征的欺骗攻击的算法。一个广为人知的漏洞是，有可能使用Play-Doh，明胶和硅树脂模具制成的人造手指来欺骗各种指纹扫描仪。活跃度检测是一种反欺骗方法，可从指纹中测量生命的生理征象，确保仅捕获活着的手指以进行注册或身份验证。在这一部分中，提出了几种新颖的新的活度检测方法。概括而言，这项工作做出了三个主要贡献。 （1）建立了一个评估嘴唇/语音/面部特征多模态生物识别系统融合中欺骗脆弱性的框架。 （2）开发了一种强大的多模式生物特征识别系统，融合了面部，语音和与文本无关的嘴唇运动。 （3）基于谷值噪声，静态出汗模式和强度分析，开发了几种新颖的活力检测算法。 （摘要由UMI缩短。）

著录项

  • 作者

    Tan, Bozhao.;

  • 作者单位

    Clarkson University.;

  • 授予单位 Clarkson University.;
  • 学科 Engineering Electronics and Electrical.
  • 学位 Ph.D.
  • 年度 2009
  • 页码 134 p.
  • 总页数 134
  • 原文格式 PDF
  • 正文语种 eng
  • 中图分类 无线电电子学、电信技术;
  • 关键词

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号