Research Initiatives

Students in our dual-degree graduate programs work on a wide variety of research projects in their respective fields. Collaborating closely with faculty advisors at Carnegie Mellon, partner universities in Portugal and often working closely with leading industry partners, research is at the core of the ICTI graduate student experience. To read more, choose from the programs listed below. If you have specific questions on a research project, email the faculty member or icti-info@andrew.cmu.edu.

  • Masters student research in Human-Computer Interaction
  • Research projects, Master of Science in Information Networking (MSIN)
  • Masters student research in Information Technology-Information Security
  • Master of Software Engineering (MSE program) Studio Projects
  • Doctoral research in Computer Science
  • Doctoral research in Software Engineering

MSIN Student Research

Study on technologies for traffic aggregation on the distribution network between VoDs

Student: Eduardo Miguel Marques Carvalho
Supervisor: Francisco Fontes, Rui Aguiar

Currently there are well-established transport networks, for clearly defined service requirements, mostly geographically well established. With the introduction of high-bandwidth services on the mobile network, the transport network is expected to have increased requirements, and to have increased Quality of Service challenges. Currently, Ethernet is appearing as a potential evolutionary technology, and two different techniques are now being discussed on the network: T-MPLS and PBB-TE. Optical based technologies are also a current trend.

This project aims to:

  1. Perform a comparison of the current and future alternatives for this core networks.
  2. Define a methodology for selection of these alternatives, given aspects as coverage, customer usage, quality of service assurances, and evolution and deployment cost.
  3. Develop a simulation/evaluation tool to apply this methodology.
  4. Apply this methodology to the Portuguese scenario.

Study on technologies for wireless distribution of 3play services

Student: Bruno Miguel Almeida Lopes
Supervisor: Rui Aguiar

3Play Services are one of the most important topics to drive today’s fixed line telcos strategy. On the other hand, the trend towards mobile communication – or simply towards wireless communications – is also a dominant factor on current telecommunications. The issue with wireless communications is the multitude of technologies that exist at this moment, at the physical (WiFi, GPRS, UMTS, WiMax) and at the service level (MBMS, Meo).

This project aims to study the existing different technologies, and how these can be used for (different types of) 3play service provision in a converged network. As such the project will:

  1. Evaluate the limitations of existing wireless technologies for 3play services.
  2. Evaluate the different service models for video provision over wireless technologies.
  3. Define a concept for the integration of streamed video over wireless technologies, taking in consideration aspects as: the existing PT infrastructures at the technology and service levels; the video adaptations required for each service; the expected service usage in different technologies.

Optmization tool for VoD server placement

Student: Gonçalo Nuno Lopes Pereira
Supervisor: Amaro de Sousa

A video-on-demand system is composed by multiple video servers and a Head-End system, where the client connects for identification. Whenever a movie is selected for provision, the front-end system redirects the request to one of the video servers where it is available. Each available movie is at one or more video servers, which has a maximum number of stored movies (limited by its storage capacity) and a maximum number of movies that can be provisioned (limited by the bandwidth of its connection to the network). 

For an optimal video provision, all video servers would have a copy of each movie. However, this would lead to extreme storage capacity servers. For an optimal video server storage, all movies would exist in a single server. However, this would lead to poor video provision performance (many video requests not been satisfied). Much more interesting is if the front-end can relate the request client position on the network with the server position on the network. If the server which gives the shortest routing path to the client is selected, the overall performance of the system will be optimized with respect to the network resource utilization. A further problem is related with the problems associated with ICC servers, and their location/operation in the whole architecture. 

This work aims:

  1. To develop a tool to plan and define the optimal location and size of some of the VoD servers.
  2. Considering also the location of the ICC servers, taking in consideration the existing ICC usage patterns.
  3. Define the movie assignment and distribution algorithms deployed in the network, under which the tool will define the location of the above boxes. 

Study on technologies for code adaptation and distribution for wireless distribution

Student: Fábio Joel Cordeiro Ferreira
Supervisor: Paulo Jorge Ferreira, Diogo Gomes

3Play Services are one of the most important topics to drive today’s fixed line telcos strategy. On the other hand, the trend towards mobile communication – or simply towards wireless communications – is also a dominant factor on current telecommunications.

Wireless video streaming has bandwidth restrictions that are usually not of concern for wired systems. Furthermore, the mobility associated with streaming in mobile environments in a converged network brings added challenges on the distribution of movies by different video servers. Digital Fountain codes is a new concept where the traditional paradigm of transmitting information as an ordered stream of packets is changed to one where the user must receive a sufficient amount of packets in order to reconstruct the original information. This concept has a natural application on reliable multicast, allowing efficient video distribution across N video server by multicast streams, optimizing network resources, at the cost of more complex decoding at the server side. Furthermore, digital fountains can also be used as the final code to be delivered to the user, allowing a flexible quality.

The project thus aims to:

  1. Study different implementation alternatives for digital fountains.
  2. Analyze the performance impact of the usage of digital fountains, in terms of extra time and power required for decoding.
  3. Define a distribution architecture for the usage of digital fountains with multicast to video server synchronization, and analyze its performance.
  4. Study the usage of digital fountains for wireless video streaming. 

Development of a wireless (WiMax) access server

Student: Tiago Silva Marques Junqueira
Supervisor: Nuno Borges

3Play Services are one of the most important topics to drive today’s fixed line telcos strategy. The evolution of the fixed line business has led to a situation which requires incumbent fixed line telcos to look f or new revenue streams and ways to keep customers loyalty. On the other hand, the trend towards mobile communication – or simply towards wireless communications – is also a dominant factor on current telecommunications. 

Portugal Telecom already has video servers infrastructures, but with the development of mobile services it is to be expected that these video servers will stream also to wireless routers. This project aims to evaluate the design of a flexible multi-technology router for video streaming, able to be used in different wireless scenarios. As such, the project aims to:

  1. Evaluate device architectures for video distribution over different radio channels.
  2. Develop a concept for an architecture potentially usable with WiFi, WiMax and 3G infrastructure.
  3. Based on this concept and on already existing work, develop an interface for WiMax technologies. 

Network performance evaluation tool

Student: Inês Freitas Oliveira
Supervisor: Susana Sargento, Peter Steenkiste

3play services are putting an increasing emphasis on network performance. Today’s services mostly demand best-effort features, while new services require high bandwidth, high availability and very stringent QoS characteristics. Services like IPTV, VoIP and VoD among others, demand that the telecom operator ensures QoS in its network. Regardless of its quality, unfortunately this cannot be perfectly assured every time. It is thus essential for the operator to keep surveillance mechanisms (applications, equipments, etc) to help the network operation and to rapidly identify causes of malfunction. 

Portugal Telecom Inovação has a product dedicated to probe the functionalities and the performance of telecom operator networks, the ArQoS, which can be used for monitoring 2play services. This product has two types of probes, intrusive and non-intrusive. A central manager controls the system, also collects/receives information from the distributed probes, correlates and analyses it, interfacing with other management systems, such as NMS platforms. The time synchronization is obtained either by GPS clock, NTP or both. 

The Non-Intrusive probes may capture traffic (e.g. RTP, UDP) and signalling (e.g., SS7, SIP)

according to specific rules. The captured information is then used to analyse the several network and services performance. The distributed intrusive probes may have different kind of interfaces (GSM, UMTS, PSTN/POTS, PSTN/ISDN and IP/Ethernet, up to 4 independent interfaces per probe. These probes may test services (like telephony, http access, Video stream, etc) or they may be used to generate specific programmable probe packets or flows.

The main objective of the proposed Project is to enhance the Intrusive probes system with new probing mechanisms that minimise the impact of the probing on the network behaviour, while simultaneously giving better estimation of the real network conditions. 

The project aims to:

  1. Explore existing models in order to understand their quality both in terms of network performance and in estimation quality, focusing on the video distribution issues.
  2. Propose improvements to existing models, in order to achieve better result reliability and/or more feasible application scenarios; one aspect to look for will be the proper time for VoD synchronization, considering aspects as network load and the introduction of a new movie release in the system.
  3. Evaluate the applicability of such models to wireless networks. 

Network Management: IP-Config - Routing Visualization Tool

Student: Ricardo Nuno Ferreira Caetano
E-Mail: ricardocaetano@cmu.edu

Introduction and background
Proactive fault-recovery can increase reliability and availability by averting the impact of impending failures in distributed systems, including router infrastructures and application services atop them. The three primary bUildingblocks of proactive fault-recovery are failure prediction, root-cause analysis, and recovery mechanisms.

By finding behavior patterns that indicate instability, one can predict failures and take the necessary measurements to avoid disturbances in the telecommunications network. 

Project proposal
The propose of this project would be the study and analysis of the deployment of core networks in a real Portuguese telecom company and to analyse the traffic in order to obtain patterns that can correctly identify anomalous situations. The project work would be obtaining speCific signatures of the system's performance. Upon this analysis, the source of the problems eQuid be identified by correlating the failure predictions for various metrics across the system. 

Student work
My proposal is to develop these system's signatures based in real traffic patterns and to correlate the various scenarios in order to obtain reliable methods to detect anomalous traffic situations and to identify unstable system's components. The project would be done in strict collaboration with the Portuguese company mentioned before. 

Student background
Electrical and Computer Engineering degree -Telecommunications area, University of Coimbra, Portugal in 2003. Currently working at Portugal Telecom (main telecom operator in Portugal). Present activities are related with PT-Comunica<;6es

International Network: interconnections and international meetings: Hanoi.

C7 signalling network, VoIP relations, GSM Roaming International Carrier services. Participation in several C7PlanningMeeting 2005 at Moscow and C7PM 2006

MSIT-IS Student Research 

Failure Characterization and Prediction for Improved Dependability of Critical Information Infrastructures

Student: Tiago Filipe Rodrigues de Carvalho
E-Mail: tcarvalho@cmu.edu
Advisor: Hyong S. Kim

Proactive fault-recovery can increase reliability and availability by averting the impact of impending failures in distributed systems, including router infrastructures and application services atop them. The three primary building-blocks of proactive fault-recovery arc failure prediction, root-cause analysis, and recovery mechanisms. The first two, in particular, are challenging open problems and this project will address the first one. 

Many faults manifest in the form of increasingly unstable behavior on various performance related metrics (e.g., network traffic, CPU usage, memory usage) prior to node/system failure. This suggests that one could predict failures by finding patterns of behaviour that indicate escalating instability in such metrics' behaviors. If one observes such patterns of behavior, warnings can be issued of (likely) imminent failure. Such predictions can be useful to system administrators as early indicators of problems. Early experiences indicate quite accurate failure prediction, and often far enough in advance for proactive correction to prevent visible failure, making this a promising avenue for further exploration. 

In this project we will characterize faults in terms of network (for instance, volume of traffic, packets lost, etc.) and system performance metrics (for instance, CPU usage, memory, CPU load, etc.), using as a testbed a real distributed system at Portugal Telecom. Some currently available tools for network discovery may be used to confirm the documented architecture of the testbed. 

The first step will be to identify fault models relevant for the specified testbed, by determining types of faults and its manifestations using both network-based and system based metrics. The accuracy of these models will be tested in terms of false positives and negatives. 

The study will evolve into characterizing the early manifestations of a fault (i.e. the component of the fault model prior to the moment of failure) in order to identify possible candidates for failure prediction at the application level. Besides using the fault model, the candidates can also be chosen (with a lower accuracy in terms of false-positives) using anomaly detection with respect to templates for normal network traffic. This could lead to the detection of faults caused by malicious interventions. 

The final result should consist in the generation of reports about predictions of possible failures. These reports may comprise the characterization of the possible fault (or a characterization of the deviation in case of anomaly-based detection) in terms of timing and metrics involved, affected network nodes, possible affected systems 'and other relevant data to identify the impact of the failure. 

Operational Management using a Common Description language Framework

Student: Luis MD Costa
E-Mail: costa@cmu.edu

Systems are subject to constant management operations that change or update their function and behavior. Managing a large infrastructure poses many difficulties to meet quality, security and efficiency requirements. Adding or removing other devices in J network introduces changes that may impact the availability, quality or security of other services or the network itself. 

Most operational tasks executed in an infrastructure comprehend configuring and gathering statistical data several systems from different vendors, by distinct company's departments, using proprietary methods. 

Large organizations have thousands of Systems that require complex organizational structures and software management to support them. Until now, the software market and the standardization communities were not able to respond to this problem as a whole. As technology nurtured an unprecedented growth in bandwidth, processing capacity and distinct protocols and services, the same growth was not achieved during the same period with the management capability of the infrastructures that are supporting this growth. 

My proposal consists in developing a Common Description Language that creates an abstraction layer between management primitives and the underlying network. This will support the study of fault tolerant network operation through monitoring of control and data plane for fast detection of faults, root-cause analysis, and seamless recovery. The main components of this project are an abstraction layer to support Network misconfiguration detection, extraction of intent-based network operation, and i);)ta analysis of network operation for root-cause analysis. 

Operational optimization of security managed services in large WAN/LAN corporate networks

Student: Ricardo M. Oliveira
E-Mail: rmoliveira@cmu.edu

PT's network has a large amount of deployed network security equipments, such as firewalls and network intrusion detection systems (NIDS).These equipments require constant maintenance, but mostly their configurations are subject to changes often: opening firewall ports, allowing traffic in some segments, tweaking NIDS configuration to ensure it accounts for the new traffic, etc. When configurations being deployed have repercussions on several of these equipments simultaneously, they need to be coordinated to ensure that changes are performed correctly. Another issue is that due to the complexity and reach of some of these changes, the influence they'll have on configurations along their path is unknown at first. At the same time, these equipments suffer from the same problems as other network equipments: rules are not often revised or removed and are merely built upon the existing ones, sometimes duplicating already existing rules.

The proposal is to address these issues by designing a way to effectively determine the path affected by configurations, deciding upon which configurations are necessary in the equipments along the way and with this ensuring that existing rules are consistent along equipments in the same network. This would be done through the analysis of configuration files and possibly of network flows. The focus of this project would be mainly in the firewall part of the problem. possibly also including NIDSs time permitting.

The benefits for PT are the possible reduction of required equipments through improved configuration and knowledge of the network and the automation of what today are manual tasks.

There's an open possibility at this time of using the firewalls and NIDSs in PT's corporate network as a testbed for this project. 

Securing and Recovering Misplaced computing devices

Student: Ricardo Nuno de Pinho Coelho Conde Marques
Advisor: Paulo Verissimo

Abstract
Everyone knows that data is valuable. The more valuable it is, the more priceless it becomes to replace if it is lost or stolen. Some solutions exist that can protect data from unauthorized disclosure, and these usually resort to some form of cryptography, and backup solutions can be used to recover data if a disaster happens. However, most users do not take advantage of these solutions for several reasons, both technical and social. Tools that assist in recovering a misplaced computer exist, but they require the computer to connect to the Internet in order to be located. In addition, they are usually extra software that needs to be installed in the computer, so they also end up not being used as much as it would be desirable. This work builds on the concepts employed by these tools and solutions, and uses some additional technology available nowadays, in order to ensure confidentiality and traceability by default. A TPM is used for confidentiality of the data and a GPS/GSM module provides traceability information, without violating the user’s privacy. The integration of this extra technology comes at an additional cost, but the user is likely to be willing to pay for it, in particular when it is compared with the cost of losing data. The usage of such technology does not put an additional burden on the user, and the usability level of the whole system is not significantly changed, as most complex operations are only done once and regular operation is performed in a user-friendly and almost transparent way.

MSE Group Project

Future Hospital

Students: Ana Rita de Almeida Ferreira Rodrigues, Anselmo José Gandarão Pereira da Silva, Nuno Alexandre Martins Seixas, Jacinto Daniel Marcelino Barbeira (Students worked as a studio-team called “Sagres Team”)
Advisors: Paulo Marques, Marco Vieira (University of Coimbra), Gil Taran (Carnegie Mellon)
Company where the project was developed: Grupo Portugal Telecom 

Abstract
Most Portuguese hospitals are faced with staff looking for equipment, patients asking for directions, and people that move between unauthorized and authorized areas without clearance. These constitute problems because they lead to:

  1. Reduced staff productivity, who waste time looking for equipment, and increased inventory costs.
  2. Patients' frustration when they get lost and again reduced staff productivity due to the time spent giving directions.
  3. Security breaches.

The market offers several systems that individually address these issues but only partially. To address the problems in a global way, it is necessary to use the functionality provided by those systems and combine it to provide new functionalities. Sagres System contains the business logic responsible for the interaction of several hospitals systems (such as HIS, ERP and Tracking). It provides functionalities that require exchange of information with those systems and reasoning about the data gathered. With this approach, hospitals can maintain their systems or implement new ones, and integrate them through our system to benefit from the added functionality it will provide.

MHCI Group Project

Usability Study of the MEO Triple Play System

Students: Catarina Pereira, Emanuel Pereira, Filipa Jervis
Advisors: Larry Constantine, Monchu Chen (Portugal), Sue Fussel, Anind K. Dey (Carnegie Mellon)
Company where the project was developed: Portugal Telecom / MEO 

Abstract
In the mid of 2007 Portugal Telecom launched Meo, a new triple play service with Internet, TV and phone, with the intent to gain entry in the TV market nowadays dominated by PT Multimedia (now ZON Multimedia). This service, to gain a good market share has to compete with a higher customer service, and the need of an online management area for Meo has emerged. Our main goals with this project are to study the usability aspects of the current Meo TV service and to get insights about how to create a new useful Meo’s online management area ‐ Managing my Meo. All the results about Meo’s TV service will be reported to the Meo team so they can improve the service. To accomplish the second goal, we applied a set of usability methods in order to: first, get information about which Meo TV features to include in Managing My Meo; second, discover what current users of Internet service management areas, use and find most valuable. The usability methods we used in this research included: user observation in their homes using Meo; first‐time users’ observation performing tasks using Meo in a laboratory; a Meo’s interface usability inspection and an online Survey. For Managing my Meo, we observed users shopping online in Amazon’s web site, the Netmadeira and Sapo ADSL online client area and we performed a myTVCABO’s interface (now myZON) usability inspection. The main problems detected in Meo’s interface were: first, confusing and misleading wording, unclear instructions to use certain features in Meo, like recordings and blocking channels; second, sometimes users’ mental models are not supported by Meo because they expect to perform actions differently, such as using the TV Guide for many tasks, like searching for a TV show when this is not well supported. 

We recommend the Meo team to:

  1. First, improve mainly the recordings section since this is one of the most valued and used features in Meo; second, the Video on Demand because it has some problems that can harm the users’ experience when renting movies. For Managing My Meo, we recommend including the following functionalities in the online management area: scheduling recordings; managing channels’ subscription; invoices; displaying Meo’s TV Guide; Video on Demand section where users can rent movies to later watch in Meo; and provide an innovative client technical support that transforms the traditional support.