Provas de Agregação do Prof. Doutor José Luís Borbinha. Relatório da Unidade Curricular: "Systems Analysis and Modelling". Seminário: "Records management and digital preservation - Two facets of information management".
access_time 04 de fevereiro de 2019 às 14:00
place Anfiteatro PA-3: Piso -1 , Pavilhão de Matemática. Instituto Superior Técnico, Campus Alameda.
local_offer Prova de Agregação
person Candidato: Prof. Doutor José Luís Brinquete Borbinha
The demand for storing and analyzing large volumes of data is today on the rise as web-based enterprises introduce innovative and interactive applications, that attract more and more users on a global scale. To cope with such data volumes, data management systems have been evolving to deliver increasingly better performance and efficiency at lower costs in large-scale scenarios. A fundamental property of these systems is data consistency. In storage systems, consistency refers to how accurate, fresh and synchronized is the state of data replicas across different machines. Most of these systems, sacrifice consistency in favor of availability and performance; while others, provide strong consistency and sacrifice availability and performance. In data processing systems, dataflow and stream continuous processing, consistency refers to the completeness state of the input that is reflected in the dataflow end output within a time frame. Traditional dataflow management systems are strongly consistent by enforcing strict temporal synchronization across processing steps. In a multitude of scenarios, such model results in inefficient executions that solely cause a marginal impact on the output, with respect to a previous state. On the other hand, stream processing systems, that deal with timestamped events, tend to be looser in terms of consistency in order to sustain low latency and not overload resources, which might not be acceptable in mission critical applications. The main goal of our research is to study performance optimizations for data-intensive management systems. At the heart of these optimizations resides the tuning of data consistency. For this tuning, we take into account the semantics of data in order to trade-off consistency for performance and resource usage in data management systems. Our evaluation indicates that we can achieve substantial performance gains, namely in terms of latency, throughput, bandwidth, and resource utilization, while keeping application outputs within acceptable levels of correctness, as defined by decision makers.
access_time 19 de dezembro de 2018 às 14:00
place Sala V0.15 (Piso 0 do Pavilhão de Civil) do IST, Alameda
local_offer Prova de Doutoramento
person Candidato: Sérgio Ricardo de Oliveira Esteves Nº 54564
supervisor_account Orientador 1: Prof. Luís Manuel Antunes Veiga
supervisor_account Orientador 2: Prof. João Nuno De Oliveira e Silva
Research on Recommendation Systems (RS) demonstrates that, although increasingly complex techniques can improve results, it is not often that the authors show concern on how such techniques can be implemented on a large scale. Within this context, this thesis intends to approach the following hypothesis: on a traditional ratings-based RS implementation, we can replace the rating by a more informative value, such that, when a traditional recommendation algorithm is applied, higher accuracy results can be obtained. Thus, our goal is to achieve the effectiveness levels of current state-of-the-art approaches, while maintaining the efficiency and practical usability of the already existing, highly scalable, software frameworks.
access_time 10 de dezembro de 2018 às 14:30
place IST Alameda - Sala 0.20, Pavilhão Informática II
local_offer Prova de CAT
person Candidato: André Filipe Caldaça da Silva Carvalho N.º 76593
supervisor_account Orientador 1: Prof. Pável Pereira Calado
supervisor_account Orientador 2: Prof. João Paulo Baptista de Carvalho
Behavior change technologies, such as physical activity trackers focusing on self-regulation techniques, face two major challenges: a) they rely on people’s motivation to constantly track and examine their behavioral data, and b) they assume people’s ability to deduce patterns and identify opportunities for action. An alternative approach would be to nudge individuals towards healthy choices right when they face a decision (e.g., “should I take the stairs or the elevator?”), thus relying less on people’s ability and motivation to regulate their behaviors. This dissertation aims to explore the design space of computer-mediated nudging technologies in the context of health behaviors, and develop a software platform for the rapid prototyping, deployment and evaluation of computer-mediated nudges.With most of current persuasive technologies enacting with conscious awareness thus holding an ongoing risk of reactance (i.e. users rejecting interventions to preserve autonomy), our first inquiry focused on the use of technology to target nonconscious processes to drive behavior change. For this purpose, we investigated how subtle influences falling outside conscious awareness could instinctively prompt users to change behavior. In this dissertation, we outline the design and development of Subly, an open-source Chrome browser plugin that subliminally primes behavioral concepts through slight emphasis on words and phrases as people browse the Internet. Subly was devised as an open-source web research tool that allows third-party researchers to design and validate their own subliminal behavior change intervention, making a contribution to the exploration of subliminal priming as a behavior change strategy. We present three studies with Subly: one that identifies the threshold of subliminal perception and one that demonstrates the efficacy of Subly in a picture-selection task. While both studies revealed promising results in consonant with prior evidence of subliminal stimuli, an inquiry into it effectiveness motivating behavior change was overlooked and it feasibility was not measured in a non-controlled setting. Motivated by this, we conducted a third inquiry to assess the feasibility of Subly in the wild, where 12 participants were exposed to subliminal cues to motivate water intake while they browsed the web. We evaluated the effectiveness of two types of stimuli: positive and neutral, and explored their influence on different hydration levels. Our results were insufficient to provide significant effects yet, they support prior work indicating that when semi deprived people infer the goal more readily than participants with low deprivation. Our inquiries allowed us to observe that subliminal priming, however, is strongly influenced by individual’s selective attention to process the stimulus unconsciously. This circumstance is shape not only by the individual, but also by the cues provided, existing the possibility that the user is not exposed to a nudge when he would benefit from it (i.e. the user may not be using the browser). These constrains led to a second project and current ongoing work. Behavior economics has identified a number of heuristics and cognitive biases - errors in thinking that deviate judgments from rational decision-making – that affect decision-making and human behavior. However, while they play a key role in how we make judgments, they are often neglected during the design of persuasive technologies and there is no knowledge on how they have been used to encourage change. In a first attempt to understand how technology can be designed to influence choice, we looked at the heuristics behind the decision-making process and used this knowledge to systematically review published empirical work in the HCI field. We reviewed 113 publications from top HCI conferences (i.e. CHI, CSCW, Ubicomp, DIS, UIST, IUI and MobileHCI). The contribution to knowledge offered by this review centers on a design pattern toolkit for the design of nudging technology that gathers behavior change techniques from a range of disciplines and supports researchers in a unified way. The toolkit is a collection of seven design strategies: Facilitate, Dissuade, Pressure, Confront, Scare, Fool and Reinforce, framed around the why (i.e. why the heuristic used can bias behavior), the what (i.e. what heuristics can be tackled to nudge) and the how (i.e. in which ways the interface can be designed to tap into specific heuristics). Our future work aims to a) lay out a set of design guidelines for the development of a card-based design tool, able to assist and provide structure to the design process, and b) the development of a web platform that supports the design of nudges in the context of health behaviors. With wearables devices being the most mobile feedback display available, we will focus our attention on smartwatches interfaces. To develop the platform, we will first conduct a set of empirical studies to examine the design and effectiveness of different techniques uncovered in the review and use the applications developed as the foundation to build the platform.
access_time 26 de novembro de 2018 às 15:30
place IST Alameda, Room 0.17 Informática II Building
local_offer Prova de CAT
person Candidato: Ana Karina Caldeira Caraban N.º 85613
supervisor_account Orientador 1: Prof. Daniel Jorge Viegas Gonçalves
supervisor_account Orientador 2: Prof. Pedro Filipe Pereira Campos
supervisor_account Orientador 3: Prof. Evangelos Karapanos
Mobile devices such as smartphones are becoming the majority of computing devices due to their evolving capabilities. Currently, service providers such as financial and healthcare institutions offer services to their clients using smartphone applications (apps). Many of these apps run on Android, the most adopted mobile operating system (OS) today. Since smartphones are designed to be carried around all the time, many persons use them to store their privatedata. However, the popularity of Android and the open nature of its app marketplaces make it a prime target for malware. This situation puts data stored in smartphones in jeopardy, as it can be stealthily stolen or modified by malware that infects the device.With the increasing popularity of smartphones and the increasing amount of personal data stored on these devices, mobile device security has drawn significant attention from both industry and academia. As a result, several security mechanisms and tools such as anti-malware software have been proposed for mobile OSs to improve the privacy of private data and to mitigate someof the security risks associated with mobile devices. However, these tools and mechanisms run in the device and assume that the mobile OS is trusted, i.e., that it is part of the trusted computing base (TCB). However, current malware often disables anti-malware software when it infects a device. For mobile phones this trend started more than a decade ago with malware such as the Metal Gear Trojan and Cabir.M, and continues to this day, e.g., with HijackRAT. In this work, we use the ARM TrustZone, a security extension for ARM processors that provides a hardware-assisted isolated environment, to implement security services that are protected from malware even if the mobile OS is compromised.In this thesis, we investigate two approaches to address some of the security risks associated with Android-based devices. In the first approach, we present security services to detect intrusions in mobile devices. We design and implement services for posture assessment (which evaluates the level of trust we can have in the device), for dynamic analysis (which performs dynamic (runtime) analysis of apps using traces of Android application programming interface (API) function calls and kernel syscalls to detect apps for malware), and for authenticity detection (which provides assurance of the authenticity and integrity of apps running on mobile devices). In the second approach, we design and implement a backup and recovery system to protect mobile devices from attacks caused by ransomware attacks, system errors, etc. Finally, we develop a software framework to facilitate the development of security services for mobile devices by combining components of the above services. As proof-of-concept, we implemented a prototype for each service and made experimental evaluations using an i.MX53 development board with an ARM processor with TrustZone.
access_time 14 de novembro de 2018 às 13:30
place EDAM room (Floor 0, Mecânica II Building) IST Campus Alameda
local_offer Prova de Doutoramento
person Candidato: Sileshi Demesie Yalew Nº 79546
supervisor_account Orientador 1: Prof. Miguel Nuno Dias Alves Pupo Correia
supervisor_account Orientador 2: Prof. Seif Haridi