الفهرس | Only 14 pages are availabe for public view |
Abstract In this thesis, a secure IoT connection between clients and broker server based on iris recognition system (Authentication Server) is proposed. The trend in this work, is to use iris as uni-modal biometrics with the aids of multi-biometrics scenarios - other than the multi-modal - to develop and implement iris-based recognition system. Regarding to image pre-processing phase, a new algorithm based on masking technique (MT) to localize iris is proposed. This technique aims to isolate the iris region free of artifacts without deformation. A modified algorithm based on masking technique (MMT); to localize iris is proposed. It solves the limitation of the iris data loss, scalability, and inconsistencies factors, for capturing conditions and different resolution images. Then, in feature extraction phase, Delta-Mean (DM) and Multi-AlgorithmMean (MAM) - as two new proposed algorithms - are developed to extract iris feature vectors. In this phase, the fusion between vectors is based on the multi-algorithm scenario which executed in a parallel processing mode. Fusion for DM and MAM based features followed by the proposed reduction method reduced the feature vector size keeping higher performance. Finally, classification is adopted by using Hamming Distance (HD) and Euclidian Distance (ED) classifiers. The evaluation of the overall proposed iris system, in verification mode using different benchmark datasets, gives the convenience of robustness and reliability under different acquisition factors. Accuracy is over 99.8% with reduction in execution time (about one second), with minimum possible features. The obtained results show the performance of the proposed solution for authentication issue. It is suitable for hardware and realtime applications, as it is fast, more accurate, and economic. Thus, It is suitable for secure MQTT connection message to authenticate right client in sensitive application over IoT. |