Date of Award


Degree Name

Master of Science


Computer Science


Martine Ceberio

Second Advisor

Eric Freudenthal


In inter-connected systems, where several computers share information with each other, problems may arise when inappropriate information starts to flow through. For example, let us consider a simple scenario of a university composed of three departments: payroll, financial aid, and academic services. We know that the payroll department deals with sensitive information, such as social security numbers, dates of birth, amounts of wages, etc. The financial aid department may use information that payroll owns. Similarly, the academic department communicates with the financial aid department. An intruder can take advantage of this network connectivity and create an inappropriate flow of information across the network, leading to the so-called Cascade Vulnerability Problem (CVP).

Several approaches have been proposed to solve this problem. Among them, the approach of Bistarelli et al. is of particular interest, as they express a solution of the problem using Constraint Programming: more specifically, soft constraints. This approach not only enables the detection of vulnerable paths in a network, but also eliminates appropriate elements of the network that provoke the security leakage.

This approach, although efficient, is limited because it overlooks important criteria in the network connectivity such as electricity, bandwidth, money, besides others. Considering such features in CVP results in generating a constraint optimization problem.

In our proposed approach, we apply the approach of Bistarelli et al, using soft constraints to model, detect and solve the cascade vulnerability problem, but we extend it by using the minimum weighted hitting set approximation algorithm to deal with the connection values. This way, we are able not only to detect the CVP, but also to make the least expensive cuts in the network's connections.




Received from ProQuest

File Size

165 pages

File Format


Rights Holder

Christian Servin