Date on Master's Thesis/Doctoral Dissertation
Computer Engineering and Computer Science
Committee Co-Chair (if applicable)
Ragade, Rammohan K.
Elmaghraby, Adel S.
Cloud computing; Information storage and retrieval; Data structures (Computer science)
As of 2014, about 2.5 quintillion bytes of data are created each day, and 90% of the data in the world was created in the last two years alone. The storage of this data can be on external hard drives, on unused space in peer-to-peer (P2P) networks or using the more currently popular approach of storing in the Cloud. When the users store their data in the Cloud, the entire data is exposed to the administrators of the services who can view and possibly misuse the data. With the growing popularity and usage of Cloud storage services like Google Drive, Dropbox etc., the concerns of privacy and security are increasing. Searching for content or documents, from this distributed stored data, given the rate of data generation, is a big challenge. Information fusion is used to extract information based on the query of the user, and combine the data and learn useful information. This problem is challenging if the data sources are distributed and heterogeneous in nature where the trustworthiness of the documents may be varied. This thesis proposes two innovative solutions to resolve both of these problems. Firstly, to remedy the situation of security and privacy of stored data, we propose an innovative Social-based Distributed Data Storage and Trust based co-operative Information Fusion Framework (SDSF). The main objective is to create a framework that assists in providing a secure storage system while not overloading a single system using a P2P like approach. This framework allows the users to share storage resources among friends and acquaintances without compromising the security or privacy and enjoying all the benefits that the Cloud storage offers. The system fragments the data and encodes it to securely store it on the unused storage capacity of the data owner's friends' resources. The system thus gives a centralized control to the user over the selection of peers to store the data. Secondly, to retrieve the stored distributed data, the proposed system performs the fusion also from distributed sources. The technique uses several algorithms to ensure the correctness of the query that is used to retrieve and combine the data to improve the information fusion accuracy and efficiency for combining the heterogeneous, distributed and massive data on the Cloud for time critical operations. We demonstrate that the retrieved documents are genuine when the trust scores are also used while retrieving the data sources. The thesis makes several research contributions. First, we implement Social Storage using erasure coding. Erasure coding fragments the data, encodes it, and through introduction of redundancy resolves issues resulting from devices failures. Second, we exploit the inherent concept of trust that is embedded in social networks to determine the nodes and build a secure net-work where the fragmented data should be stored since the social network consists of a network of friends, family and acquaintances. The trust between the friends, and availability of the devices allows the user to make an informed choice about where the information should be stored using `k' optimal paths. Thirdly, for the purpose of retrieval of this distributed stored data, we propose information fusion on distributed data using a combination of Enhanced N-grams (to ensure correctness of the query), Semantic Machine Learning (to extract the documents based on the context and not just bag of words and also considering the trust score) and Map Reduce (NSM) Algorithms. Lastly we evaluate the performance of distributed storage of SDSF using era- sure coding and identify the social storage providers based on trust and evaluate their trustworthiness. We also evaluate the performance of our information fusion algorithms in distributed storage systems. Thus, the system using SDSF framework, implements the beneficial features of P2P networks and Cloud storage while avoiding the pitfalls of these systems. The multi-layered encrypting ensures that all other users, including the system administrators cannot decode the stored data. The application of NSM algorithm improves the effectiveness of fusion since large number of genuine documents are retrieved for fusion.
Polina, Phani Chakravarthy, "SDSF : social-networking trust based distributed data storage and co-operative information fusion." (2014). Electronic Theses and Dissertations. Paper 1758.