|ICEx/DCC, sala 4329, +55 (31) 3409-1488|
|2018 a Atual||Experimentos Randomizados em Redes Sociais|
A/B tests are randomized experiments frequently used by companies that offer services on the Web for assessing the impact of new features. During an experiment, each user is randomly redirected to one of two versions of the website, called treatments. Several response models were proposed to describe the behavior of a user in a social network website, where the treatment assigned to her neighbors must be taken into account. However, there is no consensus as to which model should be applied to a given dataset. In this project, we propose a new response model, derive theoretical limits for the estimation error of several models, and obtain empirical results for cases where the response model was misspecified.
Integrantes: Fabrício Murai Ferreira (coordenador).
|2018 a Atual||Artificial Intelligence for Spotting Fake Profiles and Anomalous Users' Behaviors on the Web|
The key role of the web in our society requires mechanisms to guarantee its legitimate use. Fake profiles and malicious users threat the well-working of the web. We witness the spread of fake content on social networks, disseminated at large scale to bias users' perceptions, with impacts on reputations, businesses and even on elections. Equally, infected devices are often part of botnets disrupting services and spreading malicious content. Solving these problems requires novel methodologies to cope with complex and multi-dimensional big data, for which ground truth is inherently lacking. Novel machine learning algorithms, such as deep neural networks, have been shown to operate well in problems facing similar challenges and, thus, are good candidates for the detection of abnormal profiles and behaviors. Our goal is to develop new models based on machine learning to describe the expected traffic and the behaviors of both users and devices. We will build models that combine information from multiple sources (e.g., online social networks and network measurements). The models will help to uncover fake profiles and suspicious activities, enhancing the legitimate use of the web.
Integrantes: Ana Paula Couto Silva (coordenador), Fabrício Murai Ferreira, Jussara Marques de Almeida, Flavio Figueiredo, Marco Mellia, Idilio Drago.
|2017 a Atual||ATMOSPHERE - Adaptive, Trustworthy, Manageable, Orchestrated, Secure, Privacy-assuring, Hybrid Ecosystem for REsilient Cloud Computing|
ATMOSPHERE aims at the design and development of an ecosystem comprised of a framework and a platform enabling the implementation of next generation trustworthy cloud services on top of an intercontinental hybrid and federated resource pool.
Integrantes: Wagner Meira Jr. (coordenador), Fabrício Murai Ferreira, Gisele L. Pappa, Alberto H. F. Laender, Antonio Luiz Pinho Ribeiro, Jussara Marques de Almeida, Marcos André Gonçalves, Dorgival Guedes, Renato Ferreira, Ana Paula Couto Silva, Italo Cunha, Raquel C. de Melo Minardi, Flavio Figueiredo.
|2017 a Atual||Improving inference by intelligently pruning the telco graph|
Telephone company (telco) data can be used to infer demographic traits of phone users (such as age and gender) as well as personality traits and preferences. These sorts of inferences are of paramount importance when launching targeted campaigns: think of breast cancer and prostate cancer awareness campaigns, for instance. Unfortunately, the raw telco data contains a large volume of call records originating from telemarketing departments or destined to businesses. These calls add substantial noise to the inference process and ultimately hurt the effectiveness of campaigns. The goal of this project is to design a framework for identifying and pruning phone numbers that are unlikely to correspond to individuals. We model the telco data as a dynamic network, where nodes represent phone numbers and directed edges encode calls that took place during a given time interval. This network will also be augmented with timestamped texts, geolocation and information about mobile data usage.
Integrantes: Fabrício Murai Ferreira (coordenador), Ana Paula Couto Silva, Jorge Brea.