Date of Graduation

5-2012

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Computer Science (PhD)

Degree Level

Graduate

Department

Computer Science & Computer Engineering

Advisor/Mentor

Panda, Brajendra N.

Committee Member

Beavers, M. Gordon

Second Committee Member

Thompson, Dale R.

Third Committee Member

Douglas, David E.

Keywords

Applied sciences; Information security; Insider threat; Relational database; Security

Abstract

The dissertation concentrates on addressing the factors and capabilities that enable insiders to violate systems security. It focuses on modeling the accumulative knowledge that insiders get throughout legal accesses, and it concentrates on analyzing the dependencies and constraints among data items and represents them using graph-based methods. The dissertation proposes new types of Knowledge Graphs (KGs) to represent insiders' knowledgebases. Furthermore, it introduces the Neural Dependency and Inference Graph (NDIG) and Constraints and Dependencies Graph (CDG) to demonstrate the dependencies and constraints among data items. The dissertation discusses in detail how insiders use knowledgebases and dependencies and constraints to get unauthorized knowledge. It suggests new approaches to predict and prevent the aforementioned threat. The proposed models use KGs, NDIG and CDG in analyzing the threat status, and leverage the effect of updates on the lifetimes of data items in insiders' knowledgebases to prevent the threat without affecting the availability of data items. Furthermore, the dissertation uses the aforementioned idea in ordering the operations of concurrent tasks such that write operations that update risky data items in knowledgebases are executed before the risky data items can be used in unauthorized inferences. In addition to unauthorized knowledge, the dissertation discusses how insiders can make unauthorized modifications in sensitive data items. It introduces new approaches to build Modification Graphs that demonstrate the authorized and unauthorized data items which insiders are able to update. To prevent this threat, the dissertation provides two methods, which are hiding sensitive dependencies and denying risky write requests. In addition to traditional RDBMS, the dissertation investigates insider threat in cloud relational database systems (cloud RDMS). It discusses the vulnerabilities in the cloud computing structure that may enable insiders to launch attacks. To prevent such threats, the dissertation suggests three models and addresses the advantages and limitations of each one.

To prove the correctness and the effectiveness of the proposed approaches, the dissertation uses well stated algorithms, theorems, proofs and simulations. The simulations have been executed according to various parameters that represent the different conditions and environments of executing tasks.

Download

Share

COinS