【作者】 李恒建；

【导师】 张家树；

【作者基本信息】 西南交通大学 ， 通信与信息系统， 2010， 博士

【摘要】 随着信息技术和网络技术的高速发展,信息安全显示出前所未有的重要性。生物识别技术以其特有的稳定性、唯一性和方便性,得到越来越广泛的应用。掌纹识别作为一项新兴的生物识别技术,因具有采样简单、图像信息丰富、用户接受程度高、不易伪造、受噪声干扰小等特点受到国内外研究人员的广泛关注。和其它生物特征识别系统一样,掌纹认证系统也存在健康隐私保护和安全隐患,而这一问题很少得到国内外研究者的关注。本论文围绕基于掌纹图像的安全认证技术展开研究,同时讨论了图像有损压缩对掌纹识别认证性能的影响,为掌纹图像的有损和无损压缩编码技术提供了参考。在分析总结前人工作的基础上,主要创新工作如下：1、给出安全掌纹认证系统的评价指标,分析了混沌密码在掌纹图像及其模版保护上的应用可行性。基于非线性动力学滤波器具有高的测度熵和高度并行结构的特点,设计了一种易于硬件实现的高效混沌流密码。熵编码位于图像编码的最后阶段,根据图像无损压缩的算法,利用非线性动力学滤波器这一安全混沌模型,设计了安全哥伦布和算术熵编码。基于前馈-反馈非线性动力学滤波器流密码发生器,设计一种安全的随机哥伦布编码,该算法具有较高的编解码效率和大的密钥空间。非线性动力学滤波器的参数设计灵活,据此设计一种安全的随机算术编码。相对于一般流密码,该算法很容易产生雪崩效应,因而具有较高的安全性。2、研究掌纹图像无损压缩及其安全编解码。通过分析和测试现有常用的无损图像压缩算法,包括基于变换(整数DCT实现的JPEG、JPEG2000和SP变换)、基于预测的无损压缩算法(LJPEG, CALIC和JPEG-LS)和基于字典压缩的方法(png和UHA),综合考虑,得到JPEG-LS在压缩效率和速度上是目前适合掌纹图像无损压缩算法的结论。根据JPEG-LS编码的特点,利用FFNDF产生混沌流密码,设计了一种复杂度低、对压缩性能没有影响的安全编解码算法。实验结果表明,该算法具有较高的加密效率和安全性。3、研究掌纹图像有损压缩对掌纹认证性能的影响。CompCode是目前最优的掌纹认证算法之一。研究了不同图像有损压缩编码对掌纹CompCode认证性能的影响,包括不同变换(基于分块DCT的JPEG和DWT的JPEG2000)、变换相同编码方式不同(如基于DWT的JPEG2000和SPIHT)和量化方式不同(基于标量的SPIHT和基于网格编码量化的QTCQ)。在详细分析实验数据的基础上,得出如下结论：在低比特率时,基于DWT变换的图像压缩算法对掌纹CompCode识别性能影响小于基于DCT的方法；在高比特率时,基于DCT的图像压缩方法对掌纹CompCode识别性能的影响小于基于DWT的。这一结论表明：有损压缩对掌纹CompCode影响并不总和图像率失真程度一致,传统的基于均方误差的评价有损图像压缩方式不适合评估生物特征图像。4、研究高分辨率掌纹图像的有损压缩安全编解码。高分辨率掌纹图像具有复杂的几何纹理结构,而传统的小波由于方向有限无法稀疏表示,因而也难以进行高效的压缩编码。对偶树小波变换有6个方向且具有近似平移不变性,能较好地表示掌纹图像的方向纹理特征,然而,其在表示图像时方向依然有限。DFB可获得更加灵活的图像方向表示,因此,提出了在对偶小波树变换的高频子带用DFB进一步分解的算法,并用噪声整形技术对分解得到的系数稀疏化以获得更高的非线性逼近效果。接着,对整形后小波系数重新交织,使相邻的子带系数具有类似小波的父子关系,便于SPIHT压缩编码。实验结果表明,该编码算法在较低比特率下解码图像具有较高的峰值信噪比和视觉效果。结合基于NDF的安全随机算术编码,提出了一种可伸缩的掌纹图像加密算法。SPIHT算法在重构小波域的四叉树过程中,很容易发生雪崩效应,即使当前有一比特发生错误,也将导致整个后续码流都无法正确解码,因此具有很高的安全性。安全算术编码在压缩编码系统的最后一级,能够在不改变图像编码框架的基础上增添安全功能,便于软硬件系统实现扩展。5、为解决不同的掌纹在不同应用场合存在的交叉匹配问题,提出了一种加密域匹配认证方案。为了获得较好的认证效果,采用多个方向的Gabor滤波器提取掌纹图像纹理特征,并且在匹配分数层对各个方向的匹配分数用Sum规则融合。利用双NDF产生混沌流密码加密掌纹特征,解决了掌纹模板被盗用时无法撤销和更新的问题。详细分析了在用户密码被盗的情况下,系统性能的退化情况。综合密钥存在与被盗时系统表现出的性能指标,给出了如何设置安全掌纹认证系统的阈值,并讨论了该系统的可撤销和更新能力。更多还原

【Abstract】 As the rapid development of information and network technology, information security has become increasingly important. Compared with other biometric traits, palmprint has the advantages of large palm area for feature extraction, the simplicity of data collection and high user acceptability. Therefore, palmprint recognition has attracted a lot of researchers’ attention both home and abroad.However, the palmprint authentication system itself has inherent security threats, such as biological images and the template does not have the secret of openness (Uusually, the biometric recognition algorithms are public), and there is very little study of the domestic and foreign concerns.In this paper, foucsing on the security technology of palmprint authentication, the image compression on recognition performance impact is investigated, and furthermore the conclusions are employed to conduct a technical reference on how to choose image compression methods for both loss and lossless palmprint image compression. Specific tasks are as follows:In chapter 1, a systematic analysis of palmprint image compression, encryption and palmprint template protections has been described.In chapter 2, the evaluation of security palmprint authentication system is given and the possibility of using chaotic cryptography to protect palmprint template is analyzed. Nonlinear Dynamics Filter has a very good security and characters for implementing in parallel. The nonlinear dynamic filter and its variations are given in the following applications:an efficient stream cipher design, as well as the secure arithmetic coding and Exp-colomb coding applications. Compared with the original stream ciphers, secure Exp-colomb coding method has a high coding efficiency and security arithmetic coding can easily lead to error propagation, which has a high safety.In chapter 3, lossless palmprint images compression of the secure storage problems are studied. First of all analysis and test popular lossless image compression algorithms, including those based on transformation (integer achieved JPEG and JPEG2000, as well as the SP-based transform coding method), based on predictive lossless compression algorithms (LJPEG, CALIC and JPEG-LS), dictionary-based compression methods (png and UHA). By testing, analysis and comparison, the JPEG-LS is suitable for lossless palmprint image compression is pointed out for its compression efficiency and speed. Combined with existing secure Exp-colomb encoding algorithm, a secure image scheme based on JPEG-LS is proposed. The encryption algorithm is the image obtained with the incomprehensible nature of the encryption with high efficiency and security.In chapter 4, the impact of using different lossy compression algorithms on the matching accuracy of palmprint recognition systems is investigated, including the the image transform, quantization strategies and coding methods employed in the compression process. Further, through the tests conducted on one of the state-of-the-art palmprint recognition algorithms-CompCode, the image compression on the impact of the palmprint recognition performance is discussed detailly and stated. In addition to that we have found PSNR to be not at all suited to predict the recognition performance in iris recognition systems. For high compression rates>15, the DWT-based compression methods such JPEG2000, SPIHT are well suited to be employed in iris recognition systems. In contrary, the JPEG shows an exciting results/performance.In chapter 5, to represent the highresolution palmprint image with rich orientational texture more effectively, we use directional filter banks to decompose the subbnad of Dual-tree Discrete Wavelet Transform for more and flexible orientations.Experiments show that the proposed scheme provides high efficient representations for palmprint images. Secure storage problems of the high-resolution palmprint images in the context of lossy compression are studied. Using of efficient SPIHT coding, combined with secure arithmetic coding, a scalable palmprint image encryption algorithm is proposed. The adavantage of secure arithmetic coding in the last stage is that even without changing the image coding framework, the provision of security features is very easy to extend, as well as ease of hardware extensions.In chapter 6, to address cross-matching and palmprints template case can not be removed and updated in the token-stolen case, a certified encryption domain matching of dual-factor authentication scheme is presented. Compared with the original palmprint templates, the cancelable ones can strengthen he discriminatory power of palmprints from different hands by increasing the inter-class divergence of different palms more effectively while maintaining the intra-class distance among palmprints of the same hands. At last, the matching stage is directly performed on the encryption domain in parallel to accelerate matching and to protect user’s privacy. The final score is obtained by fusing the different directional matching scores via a SUM rule, in which not only is the speed accelerated, but the performance is improved as well.更多还原