Masters Dissertations

Permanent URI for this collection

http://192.168.13.253:4000/handle/123456789/1209

Browse

Browsing Masters Dissertations by Subject "Algorithms"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Homomorphic cryptography techniques-based algorithm for enhanced privacy of users’ data in social network systems in Ubiquitous computing
    (University of Dar es Salaam, 2019) Mshangi, Maduhu
    The use of social network systems (SNS) for interaction and sharing of information across the globe has led to new vulnerabilities and cyber threats. This has resulted in privacy violation of users’ data in SNS such as Facebook, Twitter, LinkedIn and Instagram. Privacy violation of users’ data is the problematic situation in which many researchers are trying to address. The study adopted soft systems methodology compounded with design science research to guide the research process. The research methods used include heuristic evaluation inspection experiment and divide and conquer. The study assessed the effectiveness and efficiency of the existing privacy controls in SNS to ascertain the research gap status quo. It revealed that the existing privacy controls in the SNS are not adequate in addressing privacy violation in SNS. The study proposes an algorithm for enhancing the privacy of users’ data in SNS based on homomorphic cryptography techniques. The performance analysis of the proposed algorithm was carried out using Big O notation and controlled experiment. The study found that the proposed algorithm has a time complexity of and execution time of shorter than 0.094 seconds for keys size of at most 1024 bit. It was compared with other related algorithms in a similar environment; their execution times were found to be more than 0.116 seconds. Thus, the proposed solution can be adopted to enhance the privacy of users’ data in SNS. It can be extended to other services accessible via untrusted environments.
  • Thumbnail Image
    Item
    Multiframe super-resolution algorithm for improved performance of human palm vein recognition system
    (University of Dar es Salaam, 2018) Venance, Lillan
    Palm vein recognition (PVR) is an upcoming biometric technology of recognizing individuals based on the geometrical arrangement of the palm veins. The PVR system consists of four fundamental stages: image acquisition, image pre-processing, feature extraction, and matching. The perfection of image acquisition and pre- processing stages determine the overall accuracy of the system. Focusing on the pre-processing stage, classical methods fail to generate more informative vein patterns by suppressing noises and blur, and by restoring useful image features (edges, lines, and contours) from the acquired image. This weakness calls for expensive commercial PVR systems that attempt to neutralize the impacts of the pre-processing stage. This research introduces Multiframe Super-Resolution (MRS) algorithm that can improve the pre-processing stage of the classical PVR systems. This approach suppresses noise and enhances spatial resolution of an image, and ensures protection of the critical features of the image. The proposed PVR system was tested using CASIA Multi-Spectral Palm-print Image Database, which contains 600 palm vein images captured at 850nm from 100 different people. Results show that the PVR system integrated with the MSR algorithm outperforms, giving false acceptance rate (FAR) and false rejection rate (FRR) of 0.67% and 1.00%, respectively, whereas the classical PVR system generated FAR of 2.00% and 4.33%, respectively. The results provides insights on the possibility of augmenting the available PVR systems with an inexpensive image acquisition scanner by embedding MSR algorithm that allows low quality images to be captured w ithout degrading system's performance.