الفهرس 
INTRODUCTIONOnly 14 pages are availabe for public view
 |  | from | 200 |  |  |
| | | from | 200 | | |
Abstract
In the remote surveillance systems, authentication techniques play a vital role to maintain confidentiality in the security applications. Moreover, the developers exploit this field in all applications concerned the internet of things (IoT) to prevent any illegal entry to IoT networks. Biometric data has a wide interest in authentication and identification purposes because of its uniqueness and irreplaceable possibility, so biometric data has the superiority over PINs and passwords technologies. Despite that the original biometric features have a bad need to be protected because these original features cannot be substituted or altered if compromised. Cancelable biometric techniques have the ability to become the master solution of the problems facing user’s biometric information security and privacy as hacking and rigging due to the potential of storing the resultant deformed features and keeping the genuine data away the security system. Thus, it is obligatory to provide new techniques to strongly inhibit stealing as well as modifying the original information of the genuine users. Meanwhile, these techniques must satisfy the conditions of secure recognition and authentication. Various applications have a bad need to employ only one or a mixture of cancelable biometric algorithms such as: banking affairs in payment alternatives, and financial transfers, social services, foundation networks, and employers’ identities in their firms. A non-invertible cipher schemes succeed to achieve fully deformed biometric templates with highly recognized authentication levels by means of auxiliary data extracted from a physical biometric modalities. In this thesis, we focus on new hybrid cancelable biometric schemes based on modern encryption algorithms combined with irreversible transformations to generate completely deformed images with reliable recognition and authentication results. The presented cancelable biometric schemes have the ability to generate fully deformed templates, but some of those techniques failed to achieve the balance between satisfying high security levels, as well as the reliable recognition. Unfortunately, major of the previous studies had not taken the noisy environment (light spots while capture, floating, red eyes,..etc) as a factor. These disturbance factors must be studied while proceeding the simulations. Moreover, almost one or two biometric modalities were exploited in most of the related work to prove their validity. Thence, in this thesis, we propose three novel and efficient hybrid algorithms for reliable identification with a remarkable increase in privacy scales. Furthermore, eight various databases of physical biometric modalities are tested with different specifications (colored, or gray) and captured in contrastive states (brightness, flashing, blurred, background darkness,…etc). In the first part of this thesis, we introduce the encryption method based on Genetic Algorithm (GA) to exploit the interesting properties of cross-over, and mutation operations. In addition to that the biometric templates suffer from dislocating processes and several permutations of image frames i.e., RGB components for colored biometric templates, and pixels’ positions for gray biometric templates. Initially, if the biometric template is a colored feature, the RGB components are extracted. Then, each component data is separated equally into four sub-images. Afterwards, histogram analysis is employed on the four sub-images. The least uniform sub-image is selected to apply the suggested cancelable scheme based on GA. The proposed scheme results in a remarkable improvement over the other traditional cancelable schemes. Moreover, we compare our first proposal with two of the recent and efficient related works such as Optical Scanning Holography (OSH), and Double Random Phase Encoding (DRPE) to prove the superiority of our results. In the second part of this thesis, we suggest an efficient cancelable biometric framework by exploiting an irreversible hybrid encryption technique that incorporates Deoxyribonucleic, Ribonucleic Acid sequences (DNA, and RNA), as well as, the evolutionary optimization technique of GA. The proposed algorithm is firstly performed by creating several encrypted biometric traits of the original users by employing the logistic map function. Secondly, the initialized encrypted frames are transformed into one vector of a binary array. Then, the resultant encrypted bits are converted to their corresponding introns, exons, and consequently to their relevant codons. Afterwards, the output of codons is stored in lists in a cloud database. Thirdly, the stored codons are replaced by new codons generated from the RNA lists by means of the first two bits of the cipher key extracted from the genuine biometric frame, and the old codons. In addition, a comparison between our second proposed work and efficient comparative studies are established at the end of this chapter. Experimental results achieve high values of the Area under the Receiver Operating characteristic (AROC) curve, but lowest values of False Acceptance (FAR), and False Rejection (FRR) Ratios. Another perspective is the security analysis, our proposed work achieves high Number of Pixels Change Rate (NPCR), and reaches almost the optimum values of Unified Average Changing Intensity (UACI) degree. In the third part of this thesis, we propose a new cancelable biometric scheme based on efficient encryption algorithm which is utilized a hyperchaotic system and Fibonnaci Q-matrix. The suggested scheme employs a six dimension hyperchaotic framework combined with matrices transformations by Fibonacci Q-matrix to fully deform the details of the original traits over two essential stages. The first stage is represented by scrambling the pixels’ locations of the genuine templates employing the six dimensional array of hyperchaotic system. The second stage is summarized in satisfying the diffusion by employing the Fibonacci Q-matrix on the resultant transferred blocks of the permuted template. The suggested cancelable work shows a great performance over the cancelable evaluation measurements and the security evaluation tools, especially, for gray pixels’ features. At the end of this chapter comparisons are achieved to validate our third proposed study. Furthermore, comparisons between our three proposed studies are established to summarize the progress of our work. Different simulations are performed to verify the reliability of the proposed cancelable biometric schemes. Therefore, we are obligated to test our three suggestions on various features’ details under different circumstances, and with different specifications. Thus, we examine variant of biometric datasets according to four species of Face databases: Olivetti Research Laboratory face (ORL, “gray”), Face Recognition Technology (FERET, “color”), Labled Faces in the Wild (LFW, “color”), and Iran faces Craw at Aberdeen (ICA, “color”). Moreover, other physical samples are extracted from two different datasets for each Ears: Mathematical Analysis of ear Images (AMI, “color”), and Indian Institute of Technology Delhi ears (IITD, “gray”), Palmprints: CASIA, and IITD palmprints (“gray, and color”), Fingerprints: International Fingerprint Verification Competition (FVC, “color”), and CASIA fingerprints (“gray”), as well as, for Iris: CASIA-V2, and UPOL databases (“gray, and color). These samples of physical features are grouped in twenty contrastive images for each type of features that are exposed in this thesis. Furthermore, the cancelable evaluation analysis for the proposed systems is presented in terms of different seven categories in the case of noise existence, which are mentioned as following: (A) Visual inspection, (B) Probabilities of True Distribution and False Distribution (PTD and PFD), (C) Correlation scores, (D) Histogram analysis, (E) FAR, or FRR, (F) AROC curve, (G) Speed analysis. All simulation results are achieved by using MATLAB platform on different gray and colored biometric templates.