Edinburgh Napier University

  Scaling big data infrastructure using multi-cloud environment has led to the demand for highly secure, resilient and reliable data sharing method. Several variants of secret sharing scheme have been proposed but there remains a gap in knowledge on the evaluation of these methods in relation to scalability, resilience and key management as volume of files generated increase and cloud outages persist. In line with these, this thesis presents an evaluation of a method that combines data fragmentation with Shamir’s secret sharing scheme known as Fragmented Secret Share System (FSSS). It applies data fragmentation using a calculated optimum fragment size and encrypts each fragment using a 256-bit AES key length before dispersal to cloudlets, the encryption key is managed using secret sharing methods as used in cryptography.

Four experiments were performed to measure the scalability, resilience and reliability in key management. The first and second experiments evaluated scalability using defined fragment blocks and an optimum fragment size. These fragment types were used to break file of varied sizes into fragments, and then encrypted and dispersed to the cloud, and recovered when required. Both were used in combination of different secret sharing policies for key management. The third experiment tested file recovery during cloud failures, while the fourth experiment focused on efficient key management.

The contributions of this thesis are of two ways: development of evaluation frameworks to measure scalability and resilience of data sharing methods; and the provision of information on relationships between file sizes and share policies combinations. While the first aimed at providing platform to measure scalability from the point of continuous production as file size and volume increase, and resilience as the potential to continue operation despite cloud outages; the second provides experimental frameworks on the effects of file sizes and share policies on overall system performance.

The results of evaluation of FSSS with similar methods showed that the fragmentation method has less overhead costs irrespective of file sizes and the share policy combination. That the inherent challenges in secret sharing scheme can only be solved through alternative means such as combining secret sharing with other data fragmentation method. In all, the system is less of any erasure coding technique, making it difficult to detect corrupt or lost fragment during file recovery.