This is an old revision of the document!


Ms Project Topics

1) Green Cloud Computing:

In 1999, Previous researchers stated that “it is estimated that the Internet equipment consumed roughly 8% of the total electricity in the United States, with a prediction to grow to 50% within a decade.” He also stated that “in the United States it takes 1 kg of coal to produce enough energy to send 5 MB of data over Internet.” Current rate of energy use in ICT sector is dramatically increasing. The downside is that the energy waste is even more dramatic. Energy hungry computers, handheld devices and server systems work with very low efficiency in terms of energy. To overcome this problem, “green computing” is being proposed (i.e. energy efficient computer and network systems).

Based on our energy optimization knowledge gained from wireless sensor networks, where battery is the most important resource and cannot be replaced easily, we will propose new methodologies to increase the energy efficiency of cloud systems; methodologies such as sleep schedules, multilayer designs, optimal placements, topology and routing optimizations. In order to work on this problem, you should have knowledge on: optimization methods, operating systems, distributed systems, tools such as MATLAB, OPNET, (maybe CPLEX/Gurobi), cluster systems. (If you are missing some of this knowhow, don't worry, we will help you learn them soon). (Coadvised with Rabun Kosar and Yunus Donmez)

2) Human Activity Classification with Wireless Sensor Networks using Active Learning:

In the last decade, Wireless Sensor Networks (WSNs) appeared as one of the emerging technologies that combine automated sensing, embedded computing and wireless networking into tiny embedded devices. This evolution not only enabled the use of miniaturized wireless sensors ranging from simple temperature sensors to more complex sensors that can monitor health status, but also enabled the use of WSNs in various applications ranging from military applications to healthcare. Monitoring human activities with multi-modal sensors can be used in different application areas. Medical applications, home monitoring and assisted living are some of the most prominent application domains. The aim of this project is to develop a WSN system that can enable the monitoring of daily activities of its users without disturbing their daily routine. The differentiator characteristics of the project are: • processing of multi-modal sensors data and inference using machine learning, • leaning the user behaviors through methods such as active learning and supporting the user in the case of drifts from their daily routines, • enabling a lifelong-learning and user specific system

3) Heterogeneous and Small Cell Networks (HetSNets):

Parallel to recent advancements in mobile user equipments and proliferation of bandwidth-hungry multimedia applications, user data traffic requirements are increasing dramatically each passing day. Eventually, conventional cellular networks become insufficient to accommodate the offered traffic load. To increase the capacity of the network, the idea of “bringing the network closer to the end-users” adopted by the telecommunications community. As a result; low-power access points, such as picocell and femtocell, deployed indoors (homes, offices, buses, etc.) in addition to macrocells and microcells mostly operate outdoors. Thus, a new kind of network which is comprised of several different cell types emerges: Heterogeneous and Small Cell Networks (HetSNets). However, this heterogeneity raises many challenging issues that still need to be addressed such as frequent handovers, topology control, self organization, interference management and security. Main contribution of this thesis is expected to be proposals of effective techniques to solve one or more of those challenges introduced by HetSNets. For this topic; very good knowledge of MATLAB, OPNET and at least one programming language required. Familiarity with a linear and a nonlinear optimization tool along with a modeling language (preferably AMPL) will be a plus.

4)Packet Level Performance Evaluation of Optimally Designed Wireless Sensor Networks (WSNs)

There are two separate research communities in the WSN domain. Some groups formulate and solve optimal network design problems through formal techniques and commercial solvers. The other group use discrete event simulators (e.g., OPNET) for evaluating the packet level performance of the WSNs. This thesis is offered for bridging the gap between these communities. We will take the optimal solutions and evaluate their packet level performance by carefully designing the necessary experiments. If needed, we will recommend changes in the optimal WSN design problem bu modifying the objective funcion or the constraints.

5) Optimal Software Defined Network (SDN) Design

It is very likely that we will see OpenFlow based SDN more and more often. There are many optimization issues which requires a solution for the efficient oepartion of SDN and related resulting overlay networks. Optimal routing on SDN is one of them. However, there are also topological problems and flow control issues. This work will consist of the formulation and solution of a related optimization problem. Moreover, a small size prototype of the resulting solution will be run on an OpenFlow based SDN for improving our hands on experience.

Bs Project Topics

Anomaly Detection in İstanbul Traffic Using ITS Data (for groups of 2 or 3 students)

An Intelligent Transportation System (ITS) utilizes various types of sensors deployed on roadside for traffic information collection. Given the traffic information, we can understand traffic status so that traffic police and drivers can take appropriate actions afterwards. The purpose of this project is to detect anomalies within the traffic by analyzing ITS data which may be due to incidents, and to estimate the influence of anomalous events in nearby incident area such as upstream and downstream routes of the incident location. Our aim is to produce a report automatically which includes all necessary information about an incident or an anomalous event. For this project, we will obtain and analyse a long-term past traffic data of İstanbul from İstanbul Büyükşehir Belediyesi Trafik Kontrol Merkezi.

Contact: Sinan Işık

Implementation of a Virtual SDN Test Bed with IoT Devices Including Smart Watches as Traffic Generators

Although SDN becomes very popular, configuration of vendors’ hardware SDN switches is still a challenging issue because of incompatibility between switches and controllers. In order to cope with such an issue, there are software switches implemented which support OpenFlow and runs on top of available OS (mostly Linux OS) and easier to be configured. Without constructing a real environment, an SDN test bed can be constructed and evaluated through this method for the project. Besides, for evaluating daily life scenarios related to augmented reality, some IoT devices such as smart watches will be used as traffic generators.

Contact: Ahmet Cihat Baktir

Determining the number of controllers for distributed control mechanism in SDN

As the scale of network gets larger, the number of controllers needs to be higher in order to maintain the reliability. However, we should keep the number of controller as minimum as possible. This project aims to determine the minimum number of controllers in a network topology for maintaining the lower bound of reliability.

Contact: Ahmet Cihat Baktir

Controller location problem for SDN

As the scale of network gets larger, the number of controllers needs to be higher in order to maintain the reliability. Beside of the number of controllers, their locations also have an important role in this manner. An optimum controller location algorithm needs to be defined for maintaining the reliability of the networks.

Contact: Ahmet Cihat Baktir

Assignment of network devices to controllers for a maintaining the reliability

As the scale of network gets larger, the number of controllers needs to be higher in order to maintain the reliability. Also their locations have an important role for this goal. However, determining the number and location of controllers are inadequate because device assignments need to be optimal. This project aims to determine which network device needs to be assigned to which controller.

Contact: Ahmet Cihat Baktir

Implementing northbound applications for network

The scope of this project is implementing some network functionalities in SDN controller such as Firewall, load balancer etc.

Contact: Ahmet Cihat Baktir

Evaluation of handover process in SDN

SDN has a support for handover thanks to its global view of network. A handover process will be evaluated within the scope of this network with OpenNet, which is an extension of Mininet.

Contact: Ahmet Cihat Baktir

Indoor localization on handheld devices with iOS

This project focuses on the design and implementation of indoor localization algorithms on smartphones and tablets. In more detail, the aim is to achieve the already established features with the Crowi application (*) on Android platform. Consequently, the algorithm will be enhanced with further capabilities with the addition of sensor readings such as accelerometer and gyroscope. The main challenge is to overcome the limitations imposed by the iOS platform, and hence the candidates are expected to have prior experience on mobile programming.

Contact: Bilgin Koşucu

(*) https://play.google.com/store/apps/details?id=tr.edu.boun.cmpe.crowi

Extension of the user experience on Crowi platform

Crowi has reached over 300 students over a one month period and accumulated valuable WLAN data and usage statistics. Currently, these data are not available to the general public. This projects aims to expand the already established, yet unannounced, features such as WiFi heat map on the campus area. The candidates are expected to design visualizations of the WiFi data on campus buildings and rooms on a time line, among other representations of user contributions. Further extensions include the improvements of the Crowi application to increase the user friendliness and satisfaction.

Contact: Bilgin Koşucu

Advanced analysis on campus WiFi data

This project aims to analyze the data collected on the Crowi platform with the collaboration of the university computer center. Some of the targeted problems are load distribution among WLAN access points and campus wide available free networks, spatial analysis of WiFi usage and availability and improvement of the indoor localization algorithm. The candidates are expected to carry out mobile implementations and pattern recognition algorithms for the analyses.

Contact: Bilgin Koşucu

Secure Key Exchange over Infrastructure-less Delay Tolerant Networks

Infrastructure-less delay tolerant networks are variants of computer network architecture that address the technical issue if a continuous network connectivity cannot be established. Since it is hard to establish continuous communication, providing secure key exchange in an interactive manner becomes challenging issue. Therefore, non-interactive secret sharing approaches are used to overcome this issue. A candidate will responsible for implementing several multi-party non-interactive secret-sharing schemes for such networks by developing an Android Application. Based on the algorithm design, energy efficiency, performance and security of the protocol will be analyzed.

Contact Person: Orhan Ermiş - orhanermis AT gmail DOT com

Implementation of Group Key Agreement Protocol(s) in Wireless Sensor Networks

There exist various implementations of Group Key Agreement Protocols in Wireless Sensor Networks (WSN). However, there are only few real life application of such protocols were proposed. Most of the proposed key exchanges schemes in WSNs are based on the probabilistic key pre-distribution approach. In this study, the real time implementation and design of a secure conventional group key agreement protocol will be proposed by using Arduino Fio wireless sensors. Based on the algorithm design, energy efficiency, performance and security of the protocol will be analyzed.

Contact Person: Orhan Ermiş - orhanermis AT gmail DOT com


Notes for Scaring the Unwilling and Lazy:

1. CmpE 475 or equivalent is a must 476/477/523/524/58C are strongly recommended.

2. Performance evaluation and implementation projects will result in long working hours in NETLAB.

3. Willingness for hard work is a must since all of these are difficult problems.

courses/theses_offered.1443772749.txt.gz · Last modified: 2015/10/02 10:59 by bilgin
www.chimeric.de Creative Commons License Valid CSS Driven by DokuWiki do yourself a favour and use a real browser - get firefox!! Recent changes RSS feed Valid XHTML 1.0