# Graduate Program Seminars Dec 1, 2017

Marcus Silva
M.Sc. Candidate
Supervisor: Dr. Edward Brown

Distributed Consent Management under the Ethereum Blockchain

Department of Computer Science
Friday, December 1, 2017, 9:00 am Room: EN 2022

Abstarct

Different Health Information Systems are required to implement consent and privacy management mechanisms consistent with current Health Information rules and legislation. In addition to problems maintaining consistency between systems, changing technology will make existing security and privacy models outdated over time. This should reduce some of the motivation towards large
centralized health information repositories, where data can be managed consistently in a single location. Non-centralized systems will no longer have the overhead of maintaining compliance with changing security standards and technology.

An Ethereum based blockchain consent management system is proposed to eliminate the overhead for distributed health care information by providing a public consent directive ledger. All systems would have access to the ledger, making it a common component for accessing consent directives by any and all health information systems. A walkthrough of a mock user story will demonstrate the use of the common blockchain by multiple systems while maintaining consistency across data repositories and data sets.

Xiao Wang
M.Sc. Candidate
Supervisor: Dr. George Miminis

An algorithm for the computation of the Singular Value Decomposition of a product of three matrices containing inverses

Department of Computer Science
Friday, December 1, 2017, 1:00pm, Room: EN 2022

Abstract

In linear algebra, the Singular Value Decomposition (SVD) is a factorization
of a matrix A = U T ΣV , where U and V are orthogonal and Σ diagonal. It is an extremely useful factorization in mathematics but also in many areas of science, especially where very large data sets can be replaced by much smaller sets that can describe the original problem almost as accurately.

When computing the SVD of a general three matrix product, for example
A3A2−1A1, while we are given n × n matrices A1, A2 and A3, one may naively compute the matrix A = A3A2−1 A1 and then the SVD of A. Given that this computation will take place in floating point arithmetic (finite precision), it will face two numerical problems. The first is unnecessary matrix multiplication and the second and more serious, is unnecessary computation of a matrix inverse. Both operations involve an order of n3 multiplications, but more importantly they may introduce unnecessary rounding errors, especially the latter, if A2 is close to a singular matrix.

In this project we show how the SVD of A = A3A2−1A1 may be computed
without actually forming A, by working only with the given data A3, A2 and
A1.

Minhaj Shunjaruff
M.Sc. Candidate
Supervisor: Dr. K. Vidyasankar

Fog Computing : A comprehensive Study and Application Assessment

Department of Computer Science
Friday, December 1, 2017, 3:00 pm Room: EN 2022

Abstract

Internet of Things (IoT) aims to create a network of interrelated devices to exchange information. The swarm of connected devices, i.e., sensors generate real-time data that can provide meaningful, actionable insights for real-time applications. However, the purpose of real-time is defeated if the data have to be sent to Cloud and wait for the result because of the latency involved in the transportation of data.

The concept of Fog Computing, coined by Cisco, extends the cloud near to the edge of the network so that data can be gathered and analyzed in real-time. The goal of this project is to study the characteristics, requirements, design, and application of Fog Computing. Through the simulation of two use cases, we would also understand what effect fog nodes can have on latency and network utilization in comparison with cloud-only deployment.