CPSC5900 Project: LDA-based Dark Web Analysis, Fall 2007
Course description: A detailed study and formal report of a computer science topic, normally performed during the last term of work toward the degree.
Prerequisite: approval of graduate Computer
Science advisor.
Members:
Analysis of dark websites is important for developing effective combating strategies against terrorism or extremists when more and more scattered terrorist cells use the ubiquity of Internet to form a community in the virtual space with a fairly low cost. Terrorists or extremists can anonymously set up various web sites embedded in large scale public Internet, forming on-line social communities to exchange ideology, spread propaganda, recruit members and plan attacks. In this paper, we will propose a method to discover and cluster the latent topics via analyzing contents of "Dark websites". The content and data from dark websites are gathered and extracted by crawlers and exported to documents. LDA (Latent Dirichlet Allocation)-based hierarchical Bayesian algorithm is used to analyze the extracted documents so as to discover latent communities from the web sites of terrorists or extremists. Latent communities are subsets of terrorist or extremist networks, distributing over the social actor space. The connections within each discovered topic are dense, whereas the connections between the topics are sparse. In contrast to the traditional clustering technology, LDA-based analysis allows one document to be classified into different topics. By using Expectation-Maximization algorithm, a Bayesian inference is carried out to learn the distribution and classify documents into corresponding latent topics. Our analyses help to gain more insights into the structure and communities of terrorists and extremists.
CPSC590 Project: Trust-based Access Control, Fall 2006
Design: Integrate Trust into Usage Control in File Sharing
Most access control models have formal access control rules to govern the authorization of a request from a principal. Trust evaluation helps to identify a principal or behaviors of a principal in a pervasive and collaborative environment when complete information on a principal is not available. This paper integrates trust management into usage control model to make file sharing decision in an ever-changing environment. The attributes associated with a certain principal and requested objects, contexts associated with a certain request, and even behaviors of a principal can change during the collaborative file sharing environment. A variety of such mutability poses challenges in file protection when resources sharing must happen during collaboration. In order to address the challenges, we propose a framework to determine trust value of a principle of a principle and thus integrate the trust into access control to make decision on resource exchange. First, a trust value for a principal is evaluated based on both observed behaviors and peer recommendations. Second, the usage-based access control rules are checked to decide the authorization of a request. Our system is dynamic because untrusted principal can be disenrolled and on-going access can be revoked when it does not meet the access control rules due to mutability. We apply our trust based-usage control framework into an application of file sharing by simulation.
Motivating examples:
Members: Chang Phuong and Li Yang
Publication: Li Yang,
Chang Phuong,
CEN 4105: Software Design and Development Project done at FIU